Instruction/ maintenance manual of the product RELEASE 7.3 IBM
Go to page of 388
HPSS Management Guide High Performanc e Storage System Release 7.3 November 2009 (Revision 1.0) HPSS Management Guide Novem ber 2009 Release 7.3 (Revision 1.
© Copy right (C) 1992, 2009 International Business Machines Corporation, The Regent s of the University of California, Los Alamos National Security, LL C, Lawr ence Liver more National Security, L L C, Sandia Corporation, and UT-Battelle. All rig hts reserved.
Ta ble o f Con te nt s Chap ter 1. HPS S 7.1 Co nfiguration Overvie w............................................................................... ... ...15 1.1. In trod uction.........................................................................
3.3.3.1. login.conf .................................................................................................................. .... ... ... ... ... .... ..41 3.3.3.2. krb5.conf (For Use with Kerberos Authe ntication On ly )....................
4.2.4. Modifying a Storage Subsy stem ............................................................................................. .... ... ... ... ... .81 4.2.5. Deleting a Storage Subsy stem..........................................................
5.1.1.2. Migration/Purg e Server Informa tion Window .............................................................. ... ... .... ... ... 133 5.1.1.3. Mover Infor mation Window ...........................................................................
6.2.4. Deleting a Storage Hierarchy Def i nition ........................................................................... ... .... ... ... ... ... .173 6.3. C lasses of S ervice ......................................................................
8.1. A dding Storage Space .................................................................................................................. 223 8.1.1. Importing Volume s into HPSS .....................................................................
9.1. L ogg ing O verview........................................................................................................... .... ... ... .. 294 9.2. L og P olicies................................................................................
13.1. Man agin g HPS S Use rs............................................................................................................... 325 13.1.1. Adding HPSS Users.................................................................................
15.1.2. Overview of the D B2 Backup Process.......................................................................... ... ... .... ... ... ... ...357 15.1.2.1. Config uring DB 2 for Online Backup........................................................
Lis t o f Tab les Table 1 . SS M General Optio ns................................................................................................. ... ... .... ...3 9 Table 2 . HPSSGU I Specific Option s................................................
Pre fa ce Who S ho ul d R ead T his Bo ok The HPSS Managem ent Guide is in tended as a resource for HPSS administrators. For those per forming t he initial config uration for a new HPSS sy stem, Chapter 1 provides a configuratio n roadmap.
HPSS Management Guide Novem ber 2009 Release 7.3 (Revision 1.0) 14.
Cha pte r 1. HPS S 7. 1 Co nfi gur at io n Ov er vie w 1.1 . In t rod uc ti on This chapter defi nes the high-level steps ne cessary to configure, st art, and verify correct operation of a new 7.1 HPSS system, whether that sy s tem i s created from scratch or create d by conversion from a 6.
listed. Each step is requi red unles s otherwis e indic ated. Each step is discusse d in more det ail i n the referenced section. 1. Confi gure storage subsystems (Sectio n 4.2.2: Creat ing a New St orage Subsyst em on page 76 ) Subsystems can be conf igured only partial ly at this ti me.
B. Create storage resource s (Sec tion 8. 1.2: Creating Storage Resources on page 234) 4. Create additional HPSS Users (Section 13.1.1: Addi ng HPSS Users on page 325) 5. Create Filesets and Junc tions (Section 10.1: File sets & Junctions List on page 308 and Section 10.
• Verify that a Core Server and Migrati on Purge Server have be en configured for each s torage subsy ste m. • Verify that each storage subs y s tem is accessibl e by using lsjunct ions and ensuring tha t there is at least one juncti on t o the Root fileset of each subsy ste m .
chosen i f specifie d by their COS ID . • Verify that class es of se rvice with m ultipl e copies have the Retry Stage Failur es fro m Secondary Copy fl ag enabled. Fil e Fa mil ie s, F ile se ts , an d Ju nc tio ns • Verify that fil e families and fil esets a re created according to the sit e’s requirements.
• Monitor free space from the top level st orage class in each hi erarchy to verify that t he migration and purge polic y are maintaining adequate free space .
Cha pte r 2. Sec ur ity a nd Sy st em Acc es s 2.1 . Se cu rit y Ser vi ces As of rele ase 6.2, HPSS no longer uses DCE security servic es. The new approach to securi ty divides services into two APIs, known as mechanisms, each of which has m ultipl e implementati ons.
This can be "unix" or "ldap ". · <authzURL> - a string used b y the authorizati on mechanism to l ocate t he securit y data for this rea lm . This should be "unix " for UN IX author ization, and for LDAP it should be an LDAP URL used to l ocate t he entry for the securi ty realm in an LDAP directory.
2.1 .2 .3 . L DAP LDAP authorizat ion is not support ed by IBM Service Agreements. The fol lowing i nformation is provided for sites planni ng to use LDAP authoriza tion with HPSS 7.1 as a site supported feature. An option for the aut horization mechanism is to store HPSS security inf ormation i n an LDAP director y .
To creat e a ne w group, use the following comm and at the hpss_ldap_admin prompt: group create -gid <gid> -name <name> [-uuid <uuid>] If no UU ID is suppli ed, one will be genera ted.
obtained from the foreign site's administrat or. An e xam ple would be: "ldap:// theirl dapserve r.foreign.com/cn=FOREIGNREALM.FOREIGN.COM" • Deleting a trusted foreign realm To delet.
rw---dt user ${HPSS_PRINCIPAL_PVR} rw-c-dt user ${HPSS_PRINCIPAL_SSM} ------t any_other PVR: rw---dt user ${HPSS_PRINCIPAL_PVL} rw-c--t user ${HPSS_PRINCIPAL_SSM} ------t any_other SSM: rwxcidt user $.
2.4 .1 . Co nf igu ri ng /U pda ti ng a Loc at ion Pol ic y The Locat ion Polic y can be crea ted and updated using the Locati on Policy window. If the Location Policy does not exist, the fiel ds will be displ ay ed with defaul t values for a new pol icy .
Maximum Request Threads. The m a ximum num ber of concurrent client request s all owed. Advice - If the Location Server is reporti ng heavy loads, i ncrease this number. If this number is above 300, consider replic ating the Location Server on a di fferent machine.
1. Add the HPSS_RESTRICTED_USER_FILE environment variable t o /var/hpss/et c/env.conf. Set the value of this varia ble to the name of the file that will contain the li st of rest ricte d users. For example: HPSS_RESTRICTED_USER_FILE=/var/hpss/ etc/re strict ed_users 2.
Fie ld D es crip ti on s Restricted Users li st This is the main por tion of the window whic h displays various infor m ation about each restri cted user. User Name . The nam e of the user that is restrict ed f rom HPSS access. Realm Name . The name of the HPSS re alm that encompasses the restricte d use r.
Cha pte r 3. Usi ng SS M 3.1 . T he SSM Sy st em Ma nag er 3.1 .1 . St ar ti ng t he S SM Sy st em M ana ge r Before starting t he SSM System Manager (SM), review the SM key environment variable s descri bed i n the HPSS Instal lation Guide, Section 3.
To help m itigate this, when the t hread pool is full, the S y s tem Manager notifies all t he threads in the thread pool that are wait ing on list updates to return to the client as if t hey just timed out as nor m al.
port hpss gui and hpssadm cli ents must acc ess to reach the Sy stem Manage r. This task can be made a bit easier if t he Sy stem Manager RPC program number is l abeled in the portmapper.
· When you have decide d on the hpssgui comm and li ne that is best for your inst allatio n, it will probably be us eful t o put the c om mand in a shel l scri pt for the convenience of al l SSM Adm inist rators and Operators. For e xam ple, create a file called “gui” and put the foll owing in it: /opt/hpss/bin/hpssgui.
• The proper authoriz ation e ntries for the user are cr eated i n the AUTHZACL table. 3. Th e proper SSM confi guration f iles are created and inst alled. See Section 3.3.1: Configuring the Syst em Manager Authenticat ion for SSM Cli ents , Section 3.
% /opt/hpss/bin/hpssuser -add john -ssm [ adding ssm user ] 1) admin 2) operator Choose SSM security level (type a number or RETURN to cancel): > 1 [ ssm user added : admin ] After SSM users are added, removed, or modified, the Sy st em Manager will aut om atica lly discover the change when the user atte mpts to login.
Access to the hpss_server_acl program, hpss user pr ogram, to the HPSS DB2 database, and to all HPSS uti lity programs s hould be closel y guarded. If an operator had permission to run these tools, he could modif y the type of aut hority granted to anyone by SSM.
Keytabs are created for the user by the hpssuser utilit y when the krb5keytab or unixkey tab authentica tion ty pe is specifi ed. Ke y tabs may also be crea ted manuall y with the hpss_krb5_keytab or hpss_unix_keytab utilit y , as described below. 3.3 .
3.3 .3 . SS M Co nfi gu rat io n F ile The hpssgui and hpssadm scr ipts use the SSM configuration file, ssm .conf for configura tion. The mkhpss util ity will create the SSM configuration file for the securit y m echanism support ed by SSM. The mkhpss utilit y will store the generated ssm.
File Option Comm and Line Option Functionali ty HPSS_SSM_SM_HOST_NAME -h Sy stem manager hostna me HPSS_SSM_USER_PREF_PATH -i Path to ssm preferences JAVA_BIN -j Path to java bin director y KRB5_CONFIG -k Ful l path to krb5.
File Option Com mand Li ne Option Functionali ty HPSS_AUTHEN_TYPE - t Authenticat or t y pe Inform ation on tuning cl ient polli ng rates for optimal performance is available in t he hpssadm and hpssgui m an pages.
Note that having encryption ty p es other than "des-cbc-crc" fi rst on t he "default _tkt_enct y pes" and "default_ tgs_enctypes" lines can cause authenti cation failures .
3.3 .5 .1 . Au to ma tic S SM Cli en t Pa ck ag ing and I ns ta lla ti on The hpssuser utili ty provides a m echanism for packaging all the necessar y client files required to execute the hpssgui program on the user's desktop host. Refer to the hpssuser m an page for m ore information on generating a n SSM Client Package.
These files m ay be instal led in any location on the SSM client machines. The user m ust have at least read access to the files. The SSM startup scripts hpssgui.pl, hpssgui.vbs, hpssadm.pl, and hpssadm.vbs provide t he user with a comm a nd line mechanism for st arting the SSM clie nt.
3.3 .6 .2 . So lu tio ns fo r O pe rat in g Th ro ugh a F irew all SSM can operat e through a firewall in thre e diffe rent wa y s : • The hpssgui and hpssad m can use ports exempted by the network administra tor as firewall exceptio ns. See the -n option describe d in t he hpssgui and hpssadm man pages.
• Verif y that the proper versi on of Java is inst alled. Add the Java bin direct ory to the user's $PATH, or use the -j s witch in the hpssgui s cript, or set JAVA_BIN in the user's ssm.conf file. Java can be downloaded from http://www.j ava.
If access through t he firewall is needed for other ports (eg., the Kerb eros kdc), set up a separate tunnel for each port the f irewall does not allow through. • On the cli ent machine, run the GUI: • For Kerber os aut henticat ion: % hpssgui.pl -S hpssgui.
applicable to the platf orm on which the graphical user inter face is running. Custom Look and Feels are also available at http:/ /www.javoot oo.com · -b "background color" • The onl y Look and Feel that supports color set tings and themes is the m etal Look and Feel.
include “static” text painted on the window background or label s on things like buttons. Text fields may appear as single or multiple lines and they may be “enterabl e” (t he displa y ed data can be alter ed) or “non-ent erable” (the displ ayed data cannot be changed directl y).
• Select/cut /copy/past e operat ions can be perfor med on enterabl e text fields; on non-enterabl e fields, only select and copy operat ions can be performed. • In som e cases, modify ing a field value or pressing a button ca uses the action to be perfor med imm ediatel y .
all of the current configurat ion’s field values. • Freeze - A checkbox that, while checked, suspends the aut om atic updates made to an SSM window. This allows reviewing information at the frozen point in ti m e. Unchecking t he checkbox will reac tivate normal update behavior.
menu item is availa ble on all SSM windows. • Edit menu - The Edit Menu i s located on all SSM data windows. From each Edit Menu, the user can access Cut , Copy and Paste functions which enable the user to remove data from text fie lds or transfer data among them .
variables m ay be overridden. 3.8 . M oni tor , O per at io ns a nd Co nf igu re M enu s Ov er vie w The Monit or, Operat ions and Configure menus are used by the Sy stem Manager to monitor, control and configure HPSS. They are avai lable onl y from the HPSS Health and Status window.
Accounting Status. Opens t he Subsyst em li st window where the Accounting Status and Start Accounting buttons can be found. Log File s Infor mation. Opens the Log Files Informati on window to display infor m ation for the HPSS log file s such as t he log file's size and state.
volume labels can be ente red and a request to add the disks to a st orage class can be submitted. • Create Tape Resources. Opens the Cre ate Tape Resources window where a list of tape volume labels can be ente red and a request to add the tapes t o a storage class can be submitted.
of the accounting pol icy . Only one accounti ng policy is allowed. • Location. Opens the Locat ion Poli cy window all owing configurati on and management of the location polic y . Only one location policy is allowed. • Logging. Opens the Logging Policies list window allowing configuration and managem ent of the logging polici es.
The HPSS Login window appears after start ing the hpssgui scr ipt. The user must supply a valid HPSS user name and password in or der to acces s SSM and monitor HPSS. If a login attempt is unsuccessful , review the user session log for an indicat ion of the proble m.
mism atched versions may cause com patibi lity probl ems. 3.9 .2 . Ab ou t HPS S The About HPSS window displ ays version i nformation and a portion of the HPSS copy right state m ent. The About HPSS window is accessibl e by selec ting the Hel p m enu's “About HPSS” submenu from any of the hpssgui windows.
When a user successful ly connects to t he Sy stem Manager through the Login window, the HPSS Heal th and Status window r eplaces the Login window on the screen. The HPSS Health and Status window will remain on the screen unti l the user exi ts or logs out.
icon is red, the client’s connection to the Sy ste m Manager is lost; when it is green, the connect ion is active. 3.9 .3 .2 . HP SS S ta tu s On the upper section of the HPSS Health and Status window are four status fields that repr esent the aggregate status of the HPSS system.
In addition t o the text which descr ibes the status, these fields are displa y ed with colore d icons. The ic on color depic ts that stat us as fol lows: • Red - Major and Criti cal probl ems • M.
HPSS Statisti cs fields show general trends i n HPSS operation s; the numbers are not all-inclusi ve. Some values may fluctuate up and down as servers ar e started or shut down. Some values, such as Byt es Moved , can be reset to zero in i ndividual Mover s and by SSM users.
user the abilit y to hide or displa y elem ents of the HPSS Health and Status window in order to opti m ize the viewable area. Under the View Menu the re is a menu item and checkbox for each window element that can be hidden.
Fie ld D es crip ti on s Start Ti me . The time the Sy s tem Mana ger was start ed. Uptime . The e lapsed cl ock time since the System Manager was started.
CPU Time. The amount of CPU time that the System Manager has consumed. Memory Usage. The amount of mem ory that the System Manager is curre ntly occupying. Process ID. The process i d of the System Manager . Hostname. The nam e of t he host where the S y s tem Manager is running.
Thread Pool Size and/or Request Queue Si ze to help with the System Manager perfor mance. However, increasing these 2 parameters could cause the System Manager to require m ore mem or y . • Data Change Notif icati ons. The number of data change not ificati ons rec eived from servers.
• Hostname. The name of the host where t he client is runni ng. • Connections. The num ber of RPC connectio ns this client has to the Sy ste m Manager. • Start Ti me. The time that the client connecte d to the Sy ste m Manager. • Connect Ti me.
Fie ld D es crip ti on s User Login Name . The nam e t hat the user entere d as t he User ID when logging into HPSS. User Authority . The authorit y level that the user has in or der to perfor m operations in SSM. This can be admin or operator. Login Time.
Percent Memory Free . The ratio of free memory to total mem ory in the hpssgui process. Total Window s O pened During This Session . The number of windows creat ed during the current user session.
HPSSGU I_USER_CFG_PATH or the configura tion fi le entr y HPSS _SSM_USER_PREF_PATH. If this option is not speci fied, the default value is <clie nt node>:<user.ho me>/hpss- ssm-pre fs . The user m ust have permissions to creat e the preferences file in the directory.
Save . Save the current preference settings to the preferences file using the sam e pref erence name. Delete . Del ete t he currentl y displayed pref erence settings from the preferen ces file. Sick and Default preference settings cannot be deleted. Reload .
Cha pte r 4. Glo ba l & Sub sy st em Con fig ur at io n This chapter dis cusses two levels of system configurati on: global and storage subsy ste m . Th e global configurati on applie s to t he entire HPSS install ation while subsy ste m configurations appl y only to servers and resources allocated to s torage subsystems.
This window allows y ou to m anage the HPSS global confi guration record. Onl y one such record is permitted per HPSS insta llation. To ope n the window, on the Heal th and Status window select the Configure menu, and from there the Global menu i tem.
Storage Subsystem conf iguration. Root User ID. The UID of the use r who has root access privileges t o the HPSS namespace. This only applies if the Root Is Superuser flag is set. COS Change Stream Count. The number of background threads tha t run i n the Core Server to proce ss Class of Service change re quests.
This window lists all t he subsystems in the HPSS system and provides the abi lity to m anage these subsy ste ms. To open the window, from the Health and Status window sel ect the Configure menu, and from there select the Subsystems menu item. To creat e a ne w subsy ste m , cli ck on the Cre ate New but ton.
subsy ste m is sele cted in the list . Delete - Delete the sel ected subs y s tem(s). This button is disabl ed unless a subsystem is selected in the list. Always cont act HPSS cust omer support before deleting a subsystem definit ion. An improperly deleted subsystem can cause s erious probl ems for an HPSS system.
This window allows an administrat or to m anage the configurati on of a stor age subsystem. The Add butt on is only displayed dur ing the creati on of a new configurat ion. The Update butt on is displa y e d when an existing configuration is being modifie d.
user does not specif y a COS or any hints with the creation request. The global configurat ion speci fies a default COS for an enti re HPSS ins tallati on. Sel ecting a COS on the st orage subsy stem confi guration window all ows the gl obal value to be over ridden for a parti cular subsystem.
DB Log Monitor Interval. The Core Server will check consi stency of Database Logs and Backup Logs at the indicated interval, specified in seconds . The logs are c onsistent if both primary and backup log directori es e xist and cont ain log fil es with the same names.
Gatekeeper is alre ady configured, simply add it to y ou r stor age subsystem's configuration. However, if i t is not y e t configured, it will be necessar y to wait until Secti on 4.2.3. 4: Assign a Gatekeeper if Requi red on page 80 to add the Gatekeeper.
4.2 .3 .7 . M igra ti on and P urg e Pol ic y O ve rri de s The migration and purge polic ies cont ain two elements, the basic policy and the storage subs y s tem specific polici es. This can be seen on t he Migrat ion Policy a nd Purge Poli cy windows.
E. Issue the following SQL comm and: db2> select count(*) from nsobject The resul t of the comm and should indi cate 2 rows in this table. 3. If any of these checks gives an unexpected result, do not del ete t he subsystem. Contact HPSS customer support .
Cha pte r 5. HPS S Ser v ers Most HPSS Server adm inist ration is performed from the SSM graphical user inter face Servers l ist window . Each HPSS se rver has an entr y in this list. 5.1 . Se rv er L ist This window facilit ates managem e nt of the configur ed HPSS servers.
The serve r’s configurat ion should be caref ully reviewed to ensure that it is correct and complete. Check t he Alarms and Events window and the HPSS log fi le to vie w SSM alarm messages relat ed to conf iguration problems. This situat ion can be caused by: • A DB2 record required by the server is m issing or inaccessib le.
rate (se e the hpssgui/hpssadm man pages for more details). If a ser ver is confi gured as executable but is not runni ng, SSM will treat it as an error. Therefore, if a server is not intende d to run for an extende d period, its Executable flag s hould be unchecked.
Execute Host . The Execute Hostna me fi eld from the server's basic confi guration record. Thi s field is intended to spec ify the host nam e on which the server is s upposed to run; however, no checking is done to verify if the server is actuall y running on the specified host.
server hangs up or otherwise won't respond t o the Shutdown com mand. Force Connect - Request t he Sy stem Manage r to immediately attempt to connect to the select ed servers.
• Storage System Manager • Startup Daem on (on each host where an HPSS serve r will be executing) The fi elds of the Serve r Configura tion window are divi ded into the foll owing sections. The Basic Controls section is at t he top of the window and the other sections are on individual tabs: • Basic Controls.
● Section 5.1.3: Move r Specif ic Confi guration on page 102 ● Section 5.1.1: Physi cal Volume Reposito ry (PVR) Specific Configur ation on page 109 ● Details about a ll of the other sections on this window, which appl y to all s erver types, ar e describe d in Sec tion 5.
Fie ld D es crip ti on s Server Name . A unique descript ive name given to the server. Ensure that t he Server Name is uni que. A server’s de scriptive nam e should be meaningful to local site administrator s and ope rators, in contrast to the server ’s correspond ing UUID, which has meaning for HPSS.
Execute Hostname . This is the hostname of the node on which the s erver will execute. It m ust match the Execute Hostname of the Startup Daemon that is to manage thi s server.
5.1 .1 .2 . I nt erf ac e C ont rol s The Interfa ce Controls section of the Server Configuratio n window i s comm on to al l servers. In the example window above, the server dis played is a Core Server. Fie ld D es crip ti on s Maximum Connecti ons. The maximum num be r of cl ients that this server can service at one time.
The Securi ty Controls section of the Server Configuratio n window i s comm on to al l servers. In the example window above, the server dis played is a Core Server. Fie ld D es crip ti on s Principal Name . The name of t he principal t he server will use to a uthentica te.
Authenticator . The argument passed to the authe nticati on mechanism i ndicated by the Authenticator Type configur ation variable and used to valida te communications. If it is a key t ab, the server m ust have r ead access to the key ta b fi le. Other access perm ission s shoul d not be set on this fil e or s ecurity can be breached.
• UTIME. Core Server bitfile time modified events. • ACL_SET. Core Server access c ontrol l ist modificati on events. • CHBFID . Core Server change bitfile identi fier events. • BFSETATTRS. Cor e Server set bitfile attr ibute events. 5.1 .1 .1 .
• TRACE. If selected, Trace messages generated b y the server are sent to the log. It is recommended that this be OFF for all servers except the Mover. These messages give detailed information about program flow and are general ly of int erest only to the serve r developer.
COS Change Retry Limit , Tape Dismount Delay , Tape Hando ff Delay , PVL Max Connection Wai t , Fragment Trim Limit and Frag ment Smallest Block c an be changed i n the Core Server while the server is running by changing the value on this sc reen, updating the metadata, then re-ini tializi ng the appropriate Core Server.
The serve r uses built -in default values for these set tings, but if the environment variables can be found in the server's environment, the server uses those values. The following is a li st of the nam es of the variables and the aspect s of the server's operatio n the y control.
for both gatekeeping and account validation. If m ultipl e Gatekeepe rs are configured, then any Gatekeeper may be contacted for account validati on requests .
5.1 .4 . Lo g Clie nt Sp ec if ic Co nfi gur at io n This window controls the l ocal log sett ings that will be in effect for t he node on which this Log Client runs. Fie ld D es crip ti on Client Port. The port number for communication between the Log Client and the HPSS Servers.
Re lat ed I nf orm at io n Section 9.5: Managin g Local Logging on page 301 . 5.1 .1 . Lo g Dae mon S pe cif ic C onf igu r at ion This window controls configurati on of the log daemon and the central HPSS log. Fie ld D es crip ti on s Log File Maximum Size.
Fie ld D es crip ti on s Storage Clas s Update Interval (seconds ). The interva l that indicates how often the MPS wil l query the Core Server i n its subs y s tem t o get the la test s torage class statist ics.
and usage. The trade-off for this value is that large buffer sizes will use more sy s tem mem ory and m ay be ineffici ent for small trans fers (e.g., if the Mover buffer size is 4MB, but client requests are 512KB, the Mover will not achieve any double buffe ring benefit because the ent ire amount of the tr ansfer f its in one Mover buffer).
Range End. If this fi eld is zero, Port Range End fie ld must also be zero. Port Range End. Used in conjunc tion with Port Range Start (See above). Vali d values are zero or any TCP port number to which the Mover may bind (tha t is great er than or equal to the value of Port Range Start ).
This will cause inetd to run t he executabl e /opt/hpss /bin/hp ss_mvr_tcp under the root user ID when a connection is detect ed on port 5002. The Mover proc ess uses the /var/ hpss/etc/ m vr_ek fil e to r ead the encryption key that will be used to authent icate all connections made to this Mover.
/etc/xi netd.d file. For example, if t he encryption key in the Mover’s type specifi c confi guration is 1234567890ABCDEF , then the encr y p tion key file (/var/hpss/ etc/ek.
Tab le 1. IR IX S yst em P ara me ter s Parameter Nam e Minim um Value Parameter Descr iption semm sl 512 Maximum num ber of semaphores per set maxdm a sz 513 Maximum DMA size (required for Am pex DST.
Tab le 3. Li nu x Sy ste m Pa ra met er s Parameter Nam e Minimum Value Parameter Description SEMMSL 512 Maximum num ber of semaphores per ID SHMMAX 0x2000000 Maximum shared memory segm ent s ize (by tes ) 5.
to all nodes or unexpected results m ay occur. • The Mover m ust be built with the LFT option. This is the default option f or all Move rs. If not all Movers have been built wit h this option, cl ients m ust expl icitl y specif y a class of s ervice which is valid for a Mover supporting the local file transfer option.
Specific Configurat ion W indow on page 120, Section 5.1.1.1: 3494 PVR Specif ic Confi guration on page 111, Secti on 5.1.1. 2: AML PVR Specific Configurat ion on page 113, and Section 5.1.1.2: SCSI PVR Specific Configurat ion W indow on page 118). These se ctions pr ovide addi tional vendor-s pecific advice on PVR/robot configurat ion.
5.1 .1 .1 . 3 494 P VR Spe ci fi c C onf ig ur at ion 5.1 .1 .1 .1 . 3 494 P VR Spe ci fi c Conf ig ur at ion Win do w Fie ld D es crip ti on s Cartridge Capacity .
only if the Support Shelf Tape checkbox is selecte d. The alarm value must be 2 or greater. Dismount Delay. When Def er Dis mounts is checked, thi s value is used by the PVL to determine the number of minutes that dismounts are dela y e d after the last data access .
named /dev/l mcp0 and /dev/lmcp1 respect ively. Control connectio ns must be made prior to configurati on of the /dev/lmcpX devices or undefined errors may result. For Linux sy stems, the sym b olic librar y nam e defi ned in / etc/ibmatl .conf (e.g., 3494a) should be used.
Weight 2 * Cartridges from other j obs mounted on this drive’s controll er + Weight 3 * Units of distance from the cart ridge to the drive This m e thod ha s the effe ct of distribut ing a stri ped tape m ount across as m any controll ers as possible for the best performance.
requesting client to the AML’s storage positions, drives, and Insert/Eject units . Access configurations for clients are set in the configuration file C:DASETCCONFIG on the OS/2 PC. The client nam e can be up to 64 alpha numeric char acters in length and is case sensitive.
1. M ake sure the AMU archive management softwar e is running and the hostname is r esolved, 2. Sel ect an OS/2 window from the Desktop and change the directo ry to C:DAS, C:> cd das 3. At the prompt, t y pe t cpstart and make sure that TCP/IP gets confi gured and that the port mapper program is star ted, C:das> tcpstart 4.
Score = Weight 1 * Cartridges from this job mounted on this drive ’s controll er + Weight 2 * Cartridges from other j obs mounted on this drive’s controll er + Weight 3 * Units of distance from th.
• Support Shel f Tape. If ON, the PVR and the PVL will s upport the removal of cart ridges from the t ape librar y using the shelf _tape utilit y . Command Device. The nam e of the device that the PVR can use to send com mands to the robot. For AIX sy st ems, thi s is genera lly /dev/s mc0 .
Score = Weight 1 * Cartridges from this job mounted on this drive ’s controll er + Weight 2 * Cartridges from other j obs mounted on this drive’s controll er + Weight 3 * Units of distance from th.
• Enforce Home Location. If ON, the SCSI PVR will always tr y to dismount a mounted ca rt back to its home location. Otherwise , it will just use the first free slot. The scsi_home utilit y can be used to view and m a nipulat e the hom e l ocation values.
requested shelf tape has been checked-in. The PVR will continue checking at t his inter val until the ta pe is checked-in. This field applies only if the Support Shelf Tape c heckbox is sel ected. The retr y value must be 30 or great er. Shelf Tape Check-In Alarm .
HPSS will us e any Cartridge Acce ss Port (CAP) in the STK Robot that has a pri ority greater than zero. When it needs a CAP, HPSS wil l pick t he highest pri ority CAP that is current ly availabl e. At lea st one CAP m ust be assigned a non-zero pri ority.
Would you like Multi-Host or Single-Host testing? Enter one of the following followed by ENTER: M Multi-host testing S Single-host testing X eXit this script Enter choice: m Would you like to define t.
server are removed from the HPSS configurat ion. The steps described in this section are general guidel ines. Speci fic procedures should be worked out with the aid of HPSS t echnical support so that t he details of the system's configurati on can be considere d.
pressing the Delete button. 5.1 . M oni tor in g Se rv er I nfo rm at ion A server that is runni ng and connected to SSM wil l all ow the SSM us er to view a nd update its information. This section describes the serve r execution statuses and c onfiguration information.
• Busy - The server is bus y performing its funct ion. Most servers do not update Usage State dy namicall y , so it i s unlikely you will see this value reporte d. • Unknown - The ser ver has not reported a recognized Usage Stat e. Administra tive State .
Com munication Status: Normal However, when the server is experi encing err ors or e ncountering abnormal conditi ons, i t will change the appropriate st ates and statuses to error values, noti fy SSM of the changes, and iss ue an al arm to SSM. Refer to Section 9.
Global Database Name . The name of the global database for the HPSS system. Subsystem Database Name . The name of the databas e which contains the subs y s tem tabl es used by this Core Server. Schema Name . The name of the databa se schema. Root File set Name .
• File Deletes. Th e number of bit file delete requests processed in the Core Server si nce st artup or last reset of the stati stics. • Last Reset Time. The last time the subsystem s tatistic s were reset. If this value is 0, the stat istics have not been reset since server startup.
• Free Tape Bytes . This is an es timate based on the sum of the estimated sizes of the partial ly written and unwritt en tape volumes. It is not, and cannot be, an accurate value as the amount of data that can be written on tapes varies with individual tape volumes and data compression levels.
must be greater than zero and is used i f the Gatekee ping Site Interface returns a wait time of zero f or the create, open, or stage request being retried.
Server when the Gatekeeper is monitoring Reques ts and a client disconnec ts. • Get Monito r Types . Stati stics from the gk_GetMonit orTypes API. This API is call ed by the Core Server to figure out what types of Request s bei ng m oni tored by the Gatekeeper.
Minimum Location Map Update Ti me. The shortest time, in seconds, neede d for a location map update. Average Locati on Map Update Time. The average time, in seconds, ne eded for a location map update. Maximum Locati on Map Update Time. The longest time, in sec onds, needed for a locat ion map update.
changes made to f ields on this window are sent directly to the Mover after t he appropriat e but ton is pressed and are effe ctive immediatel y . Fie ld D es crip ti on s Server Name . The descriptive name of the Mover. Number of Request Tasks. The num ber of Mover request-process ing tasks t hat currentl y exist .
This window allows y ou to view the type-specific information associat ed with a PVL. Fie ld D es crip ti on s Server Name . The descriptive name of the PVL. Total Volumes . The total num ber of volum es that have been im ported into the PVL. Total Reposito ries .
Characterist ics. Flags for the PVR: • Defer Dismounts. If ON, the PVL will delay the dismounting of a tape cartridge until the drive is required by another j ob or unti l the Dismount Delay time limit is exceede d.
heavily used cont roller, then a m ore dist ant dri ve will be selected. Retry Mount Time Limit. Th e defa ult value for this field is -1. When the defaul t val ue (-1) is used, if an error is encountered during a PVR mount operatio n, the m ount wil l pend and be retri ed every 5 minutes.
5.1 .1 .2 .1 . AM L PVR I nf or mat io n W indo w Fie ld D es crip ti on s Server Name . The descriptive name of the PVR. Total Cartridg es . The number of cartridges currentl y being m anaged b y the PVR. Cartridge Capacity . The total number of cart ridge slot s in t he librar y dedicated to this HPSS PVR.
the number of consecutive m ount errors which occur t o any drive in thi s PVR equal or exceed this value, the drive is automatical ly locked by the PVL.
alarm. Same Job on Controller , Other Job on Controller , & Distance To Drive . T hese values are used by the PVR when sel ecting a drive for a ta pe mount operati on. The three values are essentiall y weights that are used to compute an overall score for each possibl e drive .
Shelf Tape Check-In Retry . The number of seconds the P VR will wait before asking the robot if a requested shelf tape has been checked-in. The PVR will continue checking at t his inter val until the ta pe is checked-in. This field applies only if the Support Shelf Tape c heckbox is sel ected.
the best performance. It also will t ry to pick control lers that are current ly driving a minimum num ber of tapes. So, in an environment with many tape drives per control ler, the above algori thm will minimize the load on an y one controller . The Distance To Drive helps m inimize mount times by m ounting the tape in a phy sica lly cl ose drive.
5.1 .1 .2 .1 . ST K PVR Inf or ma tio n W ind ow Fie ld D es crip ti on s Server Name . The descriptive name of the PVR. Total Cartridg es . The number of cartridges currentl y being m anaged b y the PVR. Cartridge Capacity . The total number of cart ridge s lots in the librar y dedicated to thi s HPSS PVR.
Server to set the VV Conditi on of the associat ed tape volume to DOWN. Once i n DOWN state, the volum e will no longer be available for read or write operations. For fur ther i nformation a bout the Core Server VV Condition, see Secti on 4.5.4.2: Core Ser ver Tape Vol ume Informati on Window on page 271.
HPSS adm inist rators and ope rators m ay use SSM to view the act ive RTM r equests. The RTM Summary window lis ts a summary of the curr ent RTM w requests. The RTM Detail window displays deta iled information for sel ected i RTM request s. 5.1 .1 . RT M Sum mar y Lis t Fie ld D es crip ti on s RTM Su mm ary List .
5.1 .2 . RT M Det ail The RTM Detail window displays a snapshot of the detai ls of the sel ected RTM requests from the RTM Summary List window. This may contain infor mation from multiple servers , Gatekeeper, Core and Mover. The actual data dis play ed will be different for each server ty pe a nd is displa y ed in a tree structure.
ReqId. The RTM reques t ident ifier. ReqCode. The action or operation that this re quest is currentl y executing. Exa m ples include "Mover write ", "PVL verify", "write tm" (tape mark), etc. ReqState. The stat e of the requested operation.
PVLJobId. The ID of the PVL job associated with t his request. MvrId. The ID of the Mover thi s request is currentl y waiting on. DeviceId. Th e Device thi s request is curretly wait ing on to c om plete a data m ove operati on. Segment. Th e ID of the st orage segment being operated on by this request.
GroupId. The Group ID of the user associated with this request . HostAddr. The address of the originat ing host RequestType. The type of this reques t (Open, Create or Stage). Oflag. The Open flags assoc iated with this file open request. StageFlags. Flags as sociated with thi s file stage operat ion.
To star t the Startup Daem on, use the “-d” option to rc.hpss: % su - % /opt/hpss/bin/rc.hpss -d [start] 5.2 .2 .2 . St ar t ing S SM The SSM Sy stem Manager configurati on metadata should have already been created by m khpss as par t of the t he infrastr ucture configur ation.
To star t a server , select the desi red server( s) from the Servers window and click on t he Start button. Verify the result of the request in the message a rea on the Servers window. In addition, monitor t he Alarms and Events window for the “Server Initialized” event.
window for t he “Server Ter minated” event. The HPSS Startup Daemon(s) and the SSM System M anager cannot be shut down from the Servers window. Select System Manage r f rom the Shutdown submenu of t he Operatio ns menu of the Heal th and St atus window to shut down the Syst em Manager.
comm a nd line sessions, will detect that the Sy s tem Ma nager has exi ted. Choosing this opt ion will pop up a confirmation window which allows the shutdown request to be approved or canceled.
button. Verify the result of the reque st in the st atus bar on the Servers window. In addition, monitor the Alarms and Events window for the “Server Repaired” event. Repairing a server does not correct the underl ying problem that caused the server's reported stat e to change.
Core Server Resets the COS Copy t o Disk , COS Change Retry Limit , Tape Dismount Delay , Tape Handoff Del ay , PVL Max Connection Wait , Fragment Trim Limit and Fragment Smalles t Block values to the values in the specific configurati on metadata record.
Servers tha t do not support reini tiali zation, or those that do not support reinitia lizing the settings in question, must be resta rted in order for configurati on modificati ons t o take affect. Som e groups of servers depend on consist ent configuration information to run properly.
Cha pte r 6. St ora ge C onf ig ur at ion This chapter desc ribes the procedure s f or creating, m odifying, and deleting st orage classes , hierarch ies, classes of service, m igration policies, purge polic ies, and file families. 6.1 . St or age Cla ss es This section describes the process of confi guring Storage Clas ses.
Inf or ma tio n But t ons . Migration Policy . Opens the confi guration window for the migration polic y that is configured for the selected storage class. This butt on will be disabled if no s torage clas ses are selected in the Storage Classes list or the select ed s torage class does not have a migration poli cy .
This window is used to manage disk storage cl ass confi gurations. Fie ld D es crip ti on s Storage Clas s ID. The num eric identifi er assigned to the storage class. Storage Clas s Name. The descript ive name of t he storage class. Storage Clas s Type.
Advice - Do not confi gure a migrat ion poli cy for a storage class at the lowes t leve l in a hie rarchy. If a migrat ion poli cy is added to a st orage class after files are created in the storage class, t hose file s may never be migrated. Use the m km prec util ity to correct this problem.
VVs, f ragmentation of the volumes may m ake it diff icult to fi nd space f or a new segment. Setti ng Average Number of St orage Segments to a larger value will incre ase the number of segments occupi ed by files, and decrease the segment size.
Min Storage Segment Size (MINSEG). The l ower bound for storage segment sizes created on volum es in this storage cla ss. This value is the product of the Stripe Length (SL) and the Min Multipl ier (MINMULT) . Max Multiplie r ( MAXMULT). The Max Storage Segment Size (MAXSEG) must be a power of 2 multiple of the Stripe Length (SL) .
This window is used to manage tape stora ge class configurati ons. Fie ld D es crip ti on s Storage Clas s ID. The num eric identifi er assigned to the storage class. Storage Clas s Name. The descript ive name of t he storage class. Storage Clas s Type.
If a migrat ion poli cy is added to a st orage class after files are created in the storage class, t hose file s may never be migrated. Use the m km prec util ity to correct this problem. See the m kmprec man page for more informati on. Warning Threshol d.
movement prot ocol overhead and hel ps to keep the data streams fl owing smoothly . VV Block Siz e must meet the f ollowing constraining requireme nts: • It m ust be an integer multiple of the Media Block Size.
If the tape media supports "fast locat e", and that feat ure is enabled for the tape devices, choose larger values of Seconds Between Tape Marks (SBTM) .
This window is used to define Warning and Critical threshol ds unique to a par ticular stora ge subsystem, overriding the values define d in t he disk storage c lass. The user may modify either the War ning percent, Critical percent or bot h for one or more of the l isted subsy ste m s.
Subsys Name . The nam e of the select ed stor age subsystem. Warning. The cur rent warni ng threshold value for the select ed subs y s tem. If the stor age class defaults are to be used, the text “default” will be displa y e d. Critical . The current c ritical threshol d value f or the sel ected subsy stem.
Warning volum es, Crit ical volumes or bot h for one or more of the listed subsy ste m s. Select a subsystem from the list and t hen modify the values on the lower portion of the window. When the new values have been entered, selec t Update to comm i t the changes.
the administra tor to reflect desir ed behavior. If files have been stored in a storage class wit hout a migration policy, and a migration poli cy is subsequentl y configured f or it, the files create d before the addit ion of the polic y will not be m igrated.
accessed from the HPSS Health and Status window's Configu re menu, submenu Storage Space , ite m Heirarchie s. Ref er to Secti on 3. 9.3: HPSS Healt h and Stat us on page 58. The fol lowing rules for creat ing storage hi erarchie s are enforced by the Hierar chies window: • A storage class may be used only once per hierar chy.
Conf igu ra tio n But to ns Create New . Open a Storage Hi erarchy window with default values. Configure. Open the selected storage hie rarchy configura tion for editing. One hierarch y from the list must be select ed before this button is active. Delete.
Fie ld D es crip ti on s Hierarchy ID. The ID associat ed with this hie rarchy . Any unique, positive 32-bit integer value. The default value is the last configured ID plus 1. Hierarchy Name. The descript ive name assoc iated with thi s hierarch y . The defa ult value i s “Hierarch y <ID>”.
do this will rende r the files unreada ble. It is recomm ended that HPSS Cust om er Support be called to assist with t his operati on. 6.3 . Cl as se s of S er vic e This section describes how to configur e classes of servi ce. 6.3 .1 . Cl as se s o f S er vic e W ind ow A COS can be crea ted and m anaged using the Classes of Servi ce window.
Create New . Open a Class of Service window contai ning default values f or a new class of service. Configure. Open the selected class(es) of service configurat ion(s) for edit ing.
is On Open. For all subs equently c reated COSes, the defa ult value i s the same as the most rece nt COS configured. Advice – Changing the Stage Code should be done with care. See Secti on 6. 3.3: Changing a Class of Service Definiti on on page 178 for detailed informati on on each of the choices f or Stage Code.
• Auto Stage Retry . When t his flag is turned on, and a valid secondary copy of the data exists, and a stage from the primary copy fails, HPSS will aut omaticall y retry the st age using the s econdary copy.
users do frequent appends, or if those who do can be relied upon to turn truncation off f or their own files, or if the syste m administrat or can easily identify fil es which are frequent ly appended.
have a significant impact. Turning the flag on const rains files that ar e alread y larger than t he Maximum File Size to their current size. Existi ng smaller files wil l be const rained to the Maximum File Size . Changing Minimum File Size can have an impact on COS selection.
6.3 .5 . Ch ang in g a File ' s Cla ss of Ser vi ce The Core Server provides a means to change t he class of service of a file. The Core Server moves the body of the fi le as appropriate to media i n the dest ination Class of Service, t hen allows the usual migration and purge algorithms for the new Class of Service to apply.
Both basic and subs y s tem speci fic migration policies are created and managed using the Mi gration Policies window. The basi c polic y m ust be cr eated be fore cre ating any subsystem speci fic pol icies. The fi elds in the basic polic y are displa y e d with default values.
Other Migration Policy List c olumns. The remaining columns provide the sam e i nformation that can be found in Section 6.4.2.1: Disk Migratio n Policy Configurati on on page 182 and Secti on 6.4.2.2 : Tape Migration Policy Configurat ion on page 185 windows.
This window allows an administrat or to m anage disk migration policies and thei r subsystem-specific overrides. Subsy ste m -specific polici es define migration rules to be applied on a subsystem bas is instead of using the default migration poli cy .
Last Update I nterval. The number of minutes t hat must pass since a fi le was l ast updated before it can become a candidate for m igration. Number of Migration Streams Per File Family . The num ber of migration st reams to be allocated to each fil e family .
storage class. Triggered Migration Options. There are four choices for managing m igration behavior when a storage class is running out of space and the next migration isn’t y et scheduled to occur : Migrate At Warning Threshold. A migration run should be started imm ediatel y when the storage cl ass warning thr eshold is exceeded.
This window allows an administrat or to m anage tape migration poli cies and their subsy ste m -specific overrides. Subsy ste m -specific polici es define migration rules to be applied on a subsystem bas is instead of using the default (basic) migration poli cy .
migration crit eria. This goal m ay not be att ainable if t he total size of a ll files not el igible for m igration is large. Total Migrati on Str eams.
based on indi vidual f iles rathe r than tape volumes, and is able to make second copies of files stored on tape. In this algorithm, individual fil es are sel ected for migration ba sed on their last write ti m e and the set tings in t he Migrati on Polic y .
To delet e a migration polic y , sele ct the policy from the M igration Policies list window and press the Delete butt on. If a basic polic y is selected, and the policy has sub-system specific polici.
Policy butt on after the window refreshes, enter the specifi c purge policy parameters, and press the Update button. This process can be repeat ed for each sub-sy stem. When a purge policy is added to an exi sting stor age class, the Migrat ion Purge Servers must be restar ted in order for the policy to take effect.
This window allows y ou to m anage a Purge Policy. Purge polici es are assigned to storage clas ses to tell the Migra tion Purge Server and Core Server how to free disk space oc cupied b y files which have been migrated. Purge poli cies appl y to disk storage classes only .
unaccessed (for r ead or write) for the length of time specified by this field. Start purge when space used exceeds. Purge will begin for a storage cl ass when the amount of it s space used exceeds this threshold. Used space includes any file in the stor age class, whether it has bee n migrated or not.
policy are applied to all storage cla sses which reference the pol icy . If the polic y is reread, the changes ar e only applied to the storage class and stor age subsystem for whic h the pol icy is reread. Core Ser vers are not able to reread purge polici es.
Fie ld D es crip ti on s The fi elds of the columns of this window are those of the File Family Configurat ion descr ibed in Section 6.6.1: Fil e Family Configurati on . Conf igu ra tio n But to ns Create New . Open a File Family Configu ration window with defaul t values.
HPSS Management Guide Novem ber 2009 Release 7.3 (Revision 1.0) 195.
Cha pte r 7. Dev ice an d Dri ve M ana gem en t Every disk and tape drive that is used by HPSS is controlle d by two servers. The PVL controls m ounts and dismounts (for disk devices these are logical operation s only), and the Mover c ontrols I/O.
may result in a large performance degradation. For SAN disk devices, the hpss_san3p_part utility with the -i option and device name must be run to assi gn a UUID to the disk device. This UUID should be used in the Device Name fi eld when conf iguring the disk device.
These windows all ow y ou to m anage a tape or disk device/ drive c onfiguration. Modifying/Updating the device/dr ive configura tion via the Tape Device Configur ation or Disk Device Configurati on windows is not permitted while the PVL, the associated Mover(s), or the a ssociated PVR(s) (for tape) are runni ng.
For IRIX syst ems, SCSI attached tape drives are ty pically referred to by pathnames of the form /dev/rmt/ tpsXdYns, where X is the SCSI controller number, and Y is the SCSI I D of the dri ve. Note that for Ampex DST drives, the tpsXdY nrns name should be used (indicati ng that the driv er should not att empt to r ewind the drive upon close ).
Advice - This opti on is suppor ted for 3590, 3590E, 3580, 3592, 9840, 9940, DST-312, DST-314, T10000 and GY- 8240 devices. • NO-DELAY Support (tape only) . An indicat ion of whether the device support s opening the device with no delay flag set, while al lowing tape I /O operati on after the open.
device. Without the reservation, it is possible for other hosts to int erleave SCSI commands t o the drive with those issued by HPSS. This e ffect could potenti ally lead to corruption of data.
must be a val id ID. The valid IDs can be found in the Aff inity list for any cartridge in the robot. Use the command “mtlib -l <devi ce name> -qV -V<volume name>” to obtain the Aff inity list for a cart ridge. Polli ng Interval (tape only).
This window allows y ou to view the list of confi gured Mover devices and PVL drives. It also provides a number of function buttons , which a llow certain operati ons t o be performed on devices or dr ives. The Device and Drive List Preferen ces window m ay be used to select the device/ drives whi ch will be displa y e d.
• Disabled - The device is locke d, which makes it unavail able f or use by HPSS. • Unknown - The st ate of the device is not known to SSM; this is usuall y caused by the controlling Mover being down or disconnected from SSM. Device Admin State . The cur rent administ rative state of the device, as report ed by its c ontrolli ng Mover.
• Unknown - The state of the drive is not known to SSM; this is usuall y caused by the PVL being down or disconnec ted from SSM Comment. This field provides a 128 character buffer in the P VL drive metadata which gives the administrato r the opportunity to associat e miscell aneous t ext with a device/drive .
Driv e Ad mi ni str at io n Bu tto n s This group of buttons affects select ed drives. All the buttons a re disabled unless one or more drives are selected (see figure above). Lock . Lock the sel ected drives, making them unavailable to HPSS. When a drive is locked, the PVL will no longer schedul e the PVL drive.
window in Add mode, a llowing y ou to creat e a new disk device and drive. Create Tape . This button i s always active . Clicking on it opens the Tape Device Configurati on window in Add mode, a llowing y ou to creat e a new tape device and drive. Configure.
Som e PVL drive configur ation at tributes can be updated dynamicall y using the PVL Drive Infor mation window (Secti on 7.2.2: PVL Drive Information Window on page 214). The setta ble fi elds in this window are updated dynam ical ly (i.e. saved to metadata and used by the PVL upon successful Update).
There are a number of situation s in which the PVL won't allow the device/drive to be dele ted: • If the device/drive is stil l att empting to notif y the Mover/P VR about be ing added • If t.
HPSS Management Guide Novem ber 2009 Release 7.3 (Revision 1.0) 210.
The Mover Devi ce Informat ion window reports the curr ent st atistics for the device, suc h as the workload histor y of the device si nce the startup of the cont rolling Mover. The Mover Devi ce Informat ion window can also be used to lock and unlock a m over device (not e: locki ng the Mover device generally is not helpful; see S 7.
means that the correspondi ng flag is set. • Read Enabled . An indicati on of whether the device is avail able for reading. • Write Enabled . An indicat ion of whether the device is available for writ ing. • Locate Support (tape only). An indicati on of whether t he device supports a high s peed (absolute ) position ing operati on.
• Multipl e Mover Tasks (disk only ). If ON , the Move r will al low multiple Mover tasks to access the disk device. • Reserve/Release (tape onl y ).
7.2 .2 . PV L Dr ive Inf or ma tio n W ind ow HPSS Management Guide Novem ber 2009 Release 7.3 (Revision 1.0) 214.
HPSS Management Guide Novem ber 2009 Release 7.3 (Revision 1.0) 215.
This window allows y ou to view/update the information associat ed wit h an HPSS drive. The PVL Drive Information window is ty pical ly used to lock and unlock drives since newly configured drives are loc ked by default and must be unlocked to be used.
PVR (tape only) . The descript ive name of the PVR used to control this dr ive. This field i s only meaningful for tape drives. Administra tive State . Thi s fie ld all ows y ou to m odif y the state of the drive. The options are : • Locked - Makes the drive unava ilable for HPSS requests.
• PVR/Mover Notify Pending - The PVL needs to notify the ass ociated PVR and Mover that the drive has been created or delet ed. • PVR Notif y Pending - The PVL needs to notif y the associat ed PVR that the drive has been created or delete d.
and thus availabilit y . To do this, the HPSS adm inist rator will need to: • Associate tape drives to a spec ific drive pool by confi guring the HPSS ta pe drives wit h a non-zero positive integer Drive Pool ID . • Modify the end client to di ctate that thei r read request be serviced b y tape drives from this particula r Drive Pool .
7.4 . Ch ang in g Dev ic e a nd Dr ive S ta te The administr ative state of a device or drive can be set to Unloc ked or Locked . This controls whether HPSS can access the drive. Changing the state of a device or drive c an be accomplishe d via the Devices and Drives list window.
Locking a disk drive has little effect since disks are l ogically mounted when the PVL initiali zes and are not usuall y unmounted; however, a disk drive must be in the locked st ate t o be delet ed. 7.4 .3 . Re pa iri ng the S tat e o f a Dev ice or D riv e A drive can ente r an er ror or suspect state as re ported b y the PVL, Mover, or both.
HPSS Management Guide Novem ber 2009 Release 7.3 (Revision 1.0) 222.
Cha pte r 8. Vol ume and S tor ag e Man age me nt This chapter desc ribes the procedure s f or adding, removing, m onitori ng, and managing storage space in the HPSS system. The basi c unit of storage whi ch can be added to the HPSS system is the volum e.
/var/hpss/et c/AML_Ej ectPo rt.conf and /var/hpss /etc/ AML_InsertPort.con f. The AML robot can have multiple insert and ejec t ports , which have t he capabili ty to handle differ ent m edia t y pe s.
information contained in the internal label will not m atch the side information passed to PVL in the Im port request . If the start of the OwnerID field is not “HPSS”, t hen the volume will be imported as a Foreign Label volume and the side will be set to zero.
This window allows the user to import tape volumes into t he HPSS sy ste m , m a king them known to the PVL server. To make them known to the Core Server so the y can be used b y HPSS, storage resources must then be cre ated for the volumes via the Create Tape Resources window.
a successful import and goes on to the next volume in the list. This m akes it easy to restart a part ially completed import (after fixing the cause of the error which terminated the first request) by clicking the Im port button again. There is no need to r em ove from the list the volumes which were imported successfull y.
“AB5329” “AB7329” The fi lling will not occur and an error will be displa y e d if the s pecified values would gener ate an inval id volum e labe l (e.g., one greater tha n zzz999). To speci fy a list of volumes from a fi le, creat e a fil e contai ning the name of each volume to be imported on a separa te li ne.
the end of the Volu me list. If Fill Count is greater than 1, multiple label s are generated us ing the entered label as a starti ng point. Maximum Volu mes Allow ed . The m aximum num b er of volum e label s that will fit in the Volume Li st. The value is 10,000 a nd is set by SSM.
An ANSI (non-UniTree) or HPSS label with a correc t Volume ID (the Volu m e ID on the label is a s expected by HPSS) Tape Im ported Label Writ ten, Tape Imported Tape Imported An ANSI or HPSS label wi.
This window allows the user to import disk volumes int o the HPSS system, making them known to the PVL and PVR servers. To make them known to the Core Server so the y can be used, they must be created via the Core Server's Create Disk Resourc es window.
List. To automatica lly generat e a l ist of volume names in the Volume List, set the Fi ll Count to the desir ed number of volumes. Set the Fill Increment t o the number by which e ach automatical ly generated label should di ffer from t he previous one.
Based on the Im port Type, the import r equest wil l be process ed depending on how the media is currentl y labeled. See Secti on 8.1.1. 4: Selecting Import Type for Disk Volumes on page 234 for more inf ormation on select ing the appropr iate Im port Type.
again. You may dismiss the window before com pleti on; however , com pleti on messages wi ll be dis play ed in a pop-up window. At this point y ou can begin enteri ng data for another import, or you can dismiss the window.
are copied to the volum es' metadata and become a permanent part of the defini tion of the volumes. The Core Server create s the necessary metadata struc tures for each of the new virt ual volumes. Each new virtual volum e is i m mediatel y available for use.
List. To automatica lly generat e a l ist of volume names in the Volume List, set the Fi ll Count to the desir ed number of volumes. Set the Fill Increment t o the number by which e ach automatical ly generated label should di ffer from t he previous one.
Server. VVs To Create . The number of virtual volum es to be created. This value deter m ines the number of rows in the Vol ume List table a t the bott om of the window. Opt ion al or I nfo rm at ion al Field s PVs in Each VV . The Stripe Width of the selected Storage Class .
This window is us ed to creat e disk stor age resources, disk virt ual volumes, in a Core Server. The disks must first be imported to the appropr iate PVL. The names of the volumes may be entered into the window one at a time, or a l ist of volum e names may be automaticall y generated from a single entry.
"AA0080" "AA0090" "AA0100" "AA0110" "AA0120" When an addit ion produces overflow in a col um n, numerical columns ar e carried over properl y to alphabetic colu m ns and vic e versa.
available. The def ault value is 1. If the l ist fi lls bef ore the Fill Count is exhausted, filling stops and a message is displa y ed. Fill Increment .
reuse the volumes, create new storage resources on them using the Create Res ources window. To remove them entirel y from the HPSS system, export them from the PVL using the Ex port Volumes window. 8.2 .1 .1 . Ru le s fo r Del et ing Res our ce s Volumes on whic h resources are t o be delete d must be empty.
Volume List at the bottom of the wi ndow. Any of the three entry methods m ay be repeat ed multiple times on the same window to add additiona l volum es to the li st. All three entr y m ethods or any combination of them may be used in succession on the same window.
the fil e in the File Containing Volume Lis t field. The volum e names fro m the file will be a dded to the Volume List . Fie ld D es crip ti on s File Containing Volume List . The name of an external file containing a list of volume labels to be added to the end of the Volume Li st.
8.2 .2 .1 . Ru le s fo r Ex por tin g V olu me s Tape car tridges m ay be physicall y exported from any m anaging robotic libra ry . To export a tape cartridge from HPSS, the administr ator m ust be f.
This window allows y ou to export t ape and disk volumes from the HPSS sy stem. Exporting a volume is equivalent to telli ng HPSS that the volume no longer exi sts. Bef ore volumes can be exported, the Core Server storage resources that des cribe the volumes must be delet ed using the proce dure descr ibed in Section 8.
Fill Count = 6 Fill Increment = 10 Volume Label = "AA0070" Labels automatically inserted into Volume List: "AA0070" "AA0080" "AA0090" "AA0100" "A.
Fie ld D es crip ti on s Eject Tapes After Exporting . If this checkbox is sel ected, the exporte d tape volum es will also be eject ed from the PVR. File Containing Volume List . The name of an external file containing a list of volume labels to be added to the end of the Volume Li st.
available to the storage class . The m igration and purge poli cies m a y need to be modified t o free up more space or to free up the space more frequentl y . In addition, the total storage spac e for t he storage c lass may need to be reviewed to determine whether it is suff icient to acc omm oda te the actual usage of the storage cl ass.
To sele ct a r ow, click on it with the m ouse; the selec tion will be highli ghted. Note that when you sel ect a row, y ou are selec ting a st orage class within a particular storage subsy ste m . See also the rel ated window Configured Storage Classes , desc ribed in Section 6.
appropriate when t he MPS is recycled. Possibl e value s are: • Waiting - Migration is not taking place a t this time. The sta rt of the next m igration is wait ing until criteri a speci fied in the migration polic y are m et. • Running - A m i gration is in progress.
• Suspend - If the purge state is Wai ting or Runni ng, this put s the purge into the Suspended st ate. • Resume - If the purge state is Sus pended, thi s returns it to Waiting and allows MPS to again begin scheduling purge runs. • Reread poli cy - Tells the MPS to refresh its purge poli cy information by rereading t he policy.
This window allows y ou to view and update the infor m ation associat ed wit h an active disk storage cl ass. It reports the storage space data as well a s any exceeded thresholds. The window also r eports de tailed information on the migration and purge stat us.
Fie ld D es crip ti on s Storage Clas s Name . The name assi gned to this storage cl ass. Storage Clas s ID . The num eric ID of this stora ge class. Storage Clas s Type . T he clas s of media ass igned to thi s storage c lass ( tape or disk) . Subsystem Name .
alarm to SSM. Critical Threshold . For disk storage classes, this val ue is a percent age of t otal stor age space. When the used space in this storage cl ass exceeds this percent of the tot al spac e, the MPS will send a criti cal al arm to SSM.
Suspended state. • Resume - If the migration st ate is Suspended, this returns it to Waiting and allows MPS to again begin scheduling migration runs. • Reread poli cy - Tells the MPS to refresh its migration polic y information b y rereading the polic y.
8.3 .3 . M PS Tap e St ora ge C las s I nf orm at io n This window allows y ou to view and update the infor m ation associat ed wit h an active tape stora ge class. It reports the storage space data as well a s any exceeded thresholds. The window also r eports de tailed information on the migration status.
Fie ld D es crip ti on s Storage Clas s Name . The name assi gned to this storage cl ass. Storage Clas s ID . The num eric ID of this stora ge class. Storage Clas s Type . T he clas s of media ass igned to thi s storage c lass ( tape or disk) . Subsystem Name .
administrat or. Start Ti me . The date and ti m e when the most recent migration run started. It m ay still be running. End Time . T he date and time when t he last migration run c om pleted. Total Unit s Proces sed . The amount of space in the storage cla ss which has been migrated during the current or or m ost recent migration run.
Core Server metadat a that describes the volumes. See Section 8.4.3, Repacking and Reclaimi ng Volumes . 8.4 .1 . F orc ing Migr a tio n The Migration Purge Server runs m igration periodical ly in the time inte rval specif ied in the migration policy.
Repack selects tape volum es in one of t wo way s . The administra tor can provide a list of tape volumes to repack, or repac k can sel ect volumes based on a number of sel ection criteria. If repac k is provided with a list of tape volum es to process, those volumes must be i n RW, RO, EOM or EM PTY Conditi on.
This window provides an interface for r epacking tape vir tual volumes. Fie ld D es crip ti on s Storage Clas s Name. The name of the storage clas s that will be repacked. Storage Clas s ID. The ID of the storage class that will be repacked. Subsystem ID.
which a lar ge percentage of the fi les have been deleted, may be selected by repack . If VV Space is 100, the com pariso n is not performed and selectio n of tape volumes to repack is m ade using the remaining criteria. Repack Opti ons . • Select Only Reti red VVs .
But to ns Reclaim. Press thi s button to start the rec laim utili ty program on the indica ted storage cla ss. Tape volum es that are described as EMPTY, and are not retired, will be reclai m ed. Sta tus messages ar e displa y e d on the status bar at the botto m of the window a t the st art and end of the reclaim.
imported. CS Volume . Once y ou have filled in both fields, clicking on this button will open the Core Server Disk Volume or Core Server Tape Volume window for the specified volume. This metadata i s created when the disk/ tape storage resources are suc cessfull y created.
• Unallocated. Th e volume has been successfull y imported int o the HPSS sy s tem and labeled. However, no storage r esources have been creat ed on it and it has not been allocated to a Core Server. It is therefor e not available to the sy s tem f or I/O.
HPSS Management Guide Novem ber 2009 Release 7.3 (Revision 1.0) 266.
This window allows y ou to view and update the infor m ation associat ed wit h an HPSS tape cartridge. Note that the Locati on Type fiel ds are represente d differe ntl y for certain types of robot s, for which Por t , Drive and Slot ( Unit , Panel , Row , and Column ) may each be displayed as 0.
note that this can also m ean tha t side 0 of the cartridge i s mounted. PVR Server . The descriptive name of the PVR which manages the cartridge . • Cartridge Type . The HPSS m edia ty p e corresponding to this cartridge. This cont rols which the type of drive in which the cartridge can be mounted.
• Port - The loc ation is a port num ber. [This option is currently not used.] • Drive - The locati on is a drive ID num be r. • Slot - The l ocation is a slot specific ation. The fol lowing fields are fi lled in (non-zero) based on the Locati on Type and whether or not the PVR has the infor mation: Port .
This window displa y s inf ormation about a disk volume as re presented by the Core Ser ver. Fie ld D es crip ti on s Name. The ASCII name of the first physical volum e that is a part of t he disk virtual volume. The entire virtual volum e can be referred to by this name.
volum e. • In D OWN conditi on, the volum e cannot be read, written or mounted. This conditi on can be used to make a disk unavailable to the system. Change the VV Conditi on of a disk virtual volume by selecti ng the des ired condi tion from the drop down menu and then pr essing the Update butt on.
Le ss C ommo nly Use d Da ta T ab Actual Length . The lengt h of disk virt ual volume in bytes. This length include s all of the spa ce set aside for system use. See Usable Length . PV Size. The lengt h in bytes of a phy sic al volume in thi s disk virtual volume.
Phy si ca l Vo lum es This is a table of physical volume attribute s f or the ph y s ical volumes that m a ke up this disk virtual volum e. Vol Name. The ASCII name of the phy s ical volume. Type. The m edia ty p e. Dev ID. The ID of the device t he phy sica l volume is m oun ted on.
This window displa y s inf ormation about a tape volum e as represente d by the Core Server. Fie ld D es crip ti on s Name. The ASCII name of the first physical volum e that is a part of t he tape virtual volume. Th e enti re virtual volum e can be referred to by this name.
• In RWC condition, the volume can be read and writ ten. Thi s is the normal operat ional state. • In RW condition, the volume can be read and writ ten, but new tape storage segments may not be created on the volume. • In RO condition, the volume can be read but not written.
its dat a. When data is be ing written to a volume, it wil l be shown in Allocated stat e. When the End Of Media m arker is reached dur ing a tape write operat ion, the volume will ent er EOM sta te. I f the num ber of segm ents on the volume drops to zero after reaching EOM, the st ate will change to Empty.
of time may pass between updates. Next Write Address. The next address that will be written on the tape volume, expressed as an HPSS Relative Stripe Address. Le ss C ommo nly Use d Da ta T ab File Family. The family to which the volume is assigned, if any .
VV Block Size. The virtua l volume block size. Thi s is the number of by tes written from a data str eam to an element of the str iped volume before the str eam switches to the next element of the st ripe. PV Bloc k Size . The size, i n by tes, of the m edia data block.
VV Conditi on cont rols t he availabili ty of the volume for the following act ions: • Creation of new storage segm ents • Reading of exi sting stor age segm ents • Writing of existing storage s.
can be rea d, but not written. Unlike RO conditi on, ta pes in EOM c ondition can only be changed to DOWN. EOM volumes cannot enter eit her RWC or RO condition.
This window allows y ou to change which PVR owns a set of cartridge s. Before initiat ing the request from the SSM window, the cart ridges must a lready be physical ly placed into a tape librar y managed by the new PVR. The li st of the volumes to be moved may be constructed in an y of three ways.
Fill Increment = 10 Volume Label = "AA0070" Labels automatically inserted into Volume List: "AA0070" "AA0080" "AA0090" "AA0100" "AA0110" &qu.
Fill Count . The number of cart ridge l abels to be added to the end of the li st when the Volume Label field is next modifie d. This number may be one or greate r. If the lis t fills up before the Fil l Count is exhausted, filli ng stops, and a message box is displayed (se e Maximum Volumes Allowed below).
8.6 .1 . PV L Job Que ue W ind ow This window shows all outstanding jobs in the PVL. From this window, the user can issue a request to view m ore information for a parti cular PVL job or to cance l it . Each PVL job repr esents a volume mount (or seri es of mounts for a striped disk or tape volume).
• Relabel - A cartri dge being rel abeled. • Sy nc M ount - A sy nchronous mount. • Tape Check-In - A cartridge being added to the libr ary. • Tape Check-Out - A cartridge being removed from the libra ry to be pl aced on t he shelf or vault. Status .
8.6 .2 . PV L Req ue st I nf or mat io n W indo w This window is displa yed when the Job Info button is pressed on the PVL Job Queue window. It allows y o u t o view the infor mation assoc iated with a PVL job/reque st. Fie ld D es crip ti on s Job ID .
• Aborting - The j ob is bei ng aborted. • Cartridge Wait - The jo b is waiting for another job to release a cartridge that it needs. • Com plete d - The job is completed. Once a job is completed, it no longer exis ts in the PVL jo b queue, and this window wil l no l onger receive an y updates.
• Tape Check- In • Tape Check- Out • Uncomm itte d • Unload Pendi ng Drive Pool ID. If non-z ero the drive pool id will rest rict this dr ive's scheduling to tape requests specif y i ng this value. This field is not applicabl e for disks.
will begin logging alarm s indicating that the appropriat e tape has y et to be checked-in. The frequency of these al arms is contr olled by the Shelf Tape Check-In Alarm field of t he PVR-specific configurat ion window. The reques ts are displa y e d in chronologic al order.
The Tape Mount Requests window displa y s tapes which need to be m ounted in a drive. All HPSS tape mount requests, including both roboti c and oper ator t ape mounts, will be displa y ed in the window. For operator PVRs, such mount requests m ean tha t a t ape must be mounted by hand.
Clear List . Clears the li st of m ount reques ts. Note that this does not cancel any mount requests, but j ust removes them from the list . Pending mounts will reappear in t he window as t he PVR pe riodicall y retries the mounts.
refresh their caches. 3. Use the retire utilit y program to reti re the old technolog y volum es. r et ire can accept a list of PVs to reti re, or can retir e all volum es in a storage cl ass tha t are in a specifie d Conditi on. dump_ss pvs can be used to creat e a lis t of PVs for reti re to process.
HPSS Management Guide Novem ber 2009 Release 7.3 (Revision 1.0) 293.
Cha pte r 9. Log gin g and St a tu s 9.1 . Lo gg ing O ve rv iew The purpose of logging is t o record event s of inte rest that occur in HPSS in the sequence they occur to support diagnostic research.
5. The SSM Alarms and Eve nts window (Sect ion 5.2.2.3: on page 151 ) A standard configurati on for l ogging services is usuall y set by the adm inistr ator during the HPSS sy stem configurati on.
• If no server-specific logging policy is configured for a server and no de fault logging polic y is configured, only Alarm and Event messages will be logged. 9.2 .2 . Lo gg ing P ol ic ies W ind ow This window is used to manage all the log policies in t he HPSS sy ste m .
Delete. Deletes the selec ted log poli cy (s). 9.2 .2 .1 . Log gin g Po lic y C onf ig ur at ion Wi ndo w The Logging Polic y window i s used to m anage a log polic y . When creati ng a new log polic y , the Descript ive Name fiel d will be blank, and a set of defa ult opt ions will be selected.
recommended that this alwa y s be selected. • EVENT. An informational message (e.g., subsystem init ializ ing, subsystem terminating) about a significa nt occurr ence in the system that is usuall y not an error. It is recommended that this always be se lected.
A server’s log policy can be modified to control the volum e of m essages to the chose n logging destinati ons. T y pi cal ly, during norm al operations, the l evel of logging may be decreased to only Alarm, Event, and Securi ty to reduce overhead.
9.3 .2 . V iew ing the Cen tra l Lo g (De log gin g) HPSS provides the abilit y to retrieve and examine HPSS log records a s a m eans of anal y zing the act ivity and behavior of HPSS. The retrieval and log record conversion process i s referred to as “delogging.
The inf ormation i s acquired from the HPSS Log Daemon. The Log Files Information window provides information about t he HPSS central log files. Log file information includes the st ate of each log fi.
9.5 .1 . Co nf igu ri ng Loc al Lo gg ing Opt io ns The “Log Message s To:” field in the Logging Cli ent speci fic conf iguration window (Secti on 5.1.4: Log Client Specific Configurati on on page 100 ) can be m odified to contr ol the destinat ions for the messages logged by the HPSS servers runni ng in a node.
This window displa y s a num ber of the most rece nt alar m and event messages which have been received by SSM. It also all ows y ou to view individual messages in greater detai l by selecting the message and pressing t he Alarm/Event Info button to bring up t he Alarm/Ev ent I nformation window.
This window displa y s al l the details of the al arm or event sel ected fr om the Alar ms and Events window. Fie ld D es crip ti on s ID . A sequence number assigned to the log m e ssage by SSM. Thi s ID is not use d outside of SSM. Log Type . General class of the message.
• Critical • Indeterm inate • Cleared These may be accompanied by a color status i ndicator: • (Red) - Critical or Ma jor alarm • (Yellow) - Minor or Warning alarm • None – Events a nd other alar m ty pes Error Code . The er ror code associated with t he proble m underly ing t he message.
environment variabl e HPSS_SSM_ALARMS with the des ired name of t he cache file. The default for HPSS_SSM_ALARMS is defined in hpss_env_defs.h as NULL. SSM will reve rt to the internal mem ory cache if it cannot access the specif ied cache file for any re ason.
could be used, for e xam ple, to reduce the size of each data request on a slow network. The int ernal cached Alarm and Event list is displ ay ed by the hpssadm program by means of its "alarm list" comm and. Thi s command has a "-c" opti on to spec ify how many of the most recent log messages in the inte rnal cop y to display.
Cha pte r 10 . F iles et s a nd Ju nc tio ns A fileset is a logical coll ection of fil es that can be managed as a single administrat ive unit , or more simply , a disjoi nt dir ectory tree . A filese t has two identifi ers: a human readable name, and a num eric fileset ID.
File Family. The name of the file famil y to which the filese t is assigned. If this fi eld cont ains "Not in a fam il y ", the files et has not been a ssigned to a family . Filese t ID . The ID number which ident ifies the file set. A fileset ID is displayed as two double- comm a-sepa rated unsigned int eger num bers.
10. 2. C re at ing a n HPS S Fil ese t This section provides infor m ation on how to cre ate HPSS fi lesets. Only the HPSS r oot user and SSM principal are al lowed to create filesets. In order to successf ully perform fil eset administrati on, the DB2 Helper Program must be bound.
Fie ld D es crip ti on s Filese t Name . The name to be assigned t o the fil eset. This name m ust be unique to the realm in which HPSS resi des. Filese t State . The state of the fi leset. If Read is ON, the fil eset wi ll be avai lable for reading. If Write is ON, the fileset will be available for wri ting.
10. 3. M an ag ing Ex is ti ng Fil es et s This section describes how to look up inf ormation on, m odify, or del ete fi lesets. 10. 3. 1. C or e S er ver Fil es et Inf orm at io n Wi nd ow This window allows an administrat or to view, update a nd/or delete the Core Server infor mation associated with a fileset .
the data or metadata. Changing the stat e to Destroy ed will prevent both rea ding and writi ng. Fie ld D es crip ti on s Filese t ID . The ID number which identifie s the fileset. A fileset ID is displa y e d as two double- comm a -separated unsigned int eger numbers.
UID . The User ID identif y ing t he user owning t he root node of the fileset. GID . The Group ID identi fy ing the principa l group owning the root node of the fileset . Permis sions . The UNIX-sty le permission s ass igned to the root node of the fi leset.
10. 5. C re at ing a Ju nc ti on Only the HPSS root user and SSM principal are allowed to create junctio ns. A juncti on is a name space obj ect that point s to a fileset and is similar t o a persiste nt UNIX mount point. The fi leset pointed to m ay reside in anot her subs y s tem.
10. 6. D ele ti ng a J un ct ion Juncti ons can be deleted using SSM or by using the util ity routine deljuncti on . For m or e infor mation, refer to the del junction man page. To delet e a j unction using SSM, sele ct the juncti on(s) to be delet ed from t he Filesets & Junctions Li st and press the Delet e Junct ion butt on.
Cha pte r 11 . F ile s, D ir ect o rie s a nd Ob jec ts by SO ID This chapter desc ribes two ways to displa y basic information about fil es and di rectories stored in HPSS. Starting with a full y qualified path name, you can look up a file or direc tory in t he sy stem and displa y information about it .
11. 1. 1. F ile /D ir ec to ry Inf or ma tio n W ind ow This window shows detai ls about a file or dir ectory. If a file i s displa y e d i n the File/ Directory Informati on window, a button labeled Show Bitfil e ID will appear at the bott om of the window.
Object Type . The ty pe of object being di splay ed, eit her File or Director y . Class of Service . The name of the Class of Service in which the file is stored. If the displa y e d object is a director y the value of this fi eld will be NONE. File Family .
Extended ACL . If any ACL entry other than the default ACL entries exist, then the file or director y is said to contain extended ACLs. There are three type of ACLs that could have extended ACLs: • .
Com ment x x Com posite Perms x x COSId x DataLength x x EntryCount x ExtendedACLs x x Family Id x FilesetHand le FilesetId x x FilesetRoot Id x x FilesetSt ateFlags x x FilesetT y pe x x GID x x Grou.
11. 2. O bj ect s by SO ID W ind ow To displ ay this window, select Monitor from the Health and Status window, then select Lookup HPSS Objects, and the n“ Objects by SOID . This window allows you to open an infor m ation window for an HPSS object which you speci fy by the objec t's HPSS Storage Obje ct ID (SOID).
Cha pte r 12 . Ta pe Agg re ga tio n This chapter dis cusses the foll owing operation s: • Overview of Tape Aggregation • Tape Aggregation Performance Consider ations • Configuring Tape Aggregation 12.
Migration Policy scree n. Edit any other tape aggregation related fields on that screen as needed. If MPS is running, y ou must also tell it to rer ead the Disk Migration Policy.
Cha pte r 13 . Us er A cc oun ts and A cco un ti ng 13. 1. M an ag ing HP SS U ser s After the HPSS system is up and running, the adm inistra tor m ust create the necessar y accounts for the HPSS users.
[ added unix user ] [ KADMIN_PRINC unset; using kadmin.local for Kerberos ops ] [ adding kerberos principal ] [ added kerberos principal ] HPSS/LDAP home directory: / Primary group ID#: 210 [ adding f.
13. 1. 1. 3. Add a Ke rbe ro s Use r I D The hpssuser utili ty invokes the kadmin util ity to create t he KRB pri ncipal and account. This can be done using bot h key t ab and password by specifying the -krbkey t ab option.
# hpssuser -add user1 -ftp -nohome User ID#: 300 Enter password for user1: ****** Re-enter password to verify: ****** Full name: Test User HPSS/LDAP home directory: /home/user1 Login shell: /bin/ksh P.
[ SSM user deleted ] [ deleting ldap principal ] [ deleted ldap principal ] [ deleting ftp user ] [ ftp user deleted ] [ deleting kerberos principal ] [ KADMIN_PRINC unset; using kadmin.local for Kerberos ops ] [ deleted kerberos principal ] [ deleting unix user ] 13.
# hpssuser -ssmclientpkg /tmp/ssmclientpkg.tar [ packaging ssm client ] [ creating /tmp/ssmclientpkg.tar ] ssm.conf krb5.conf hpssgui.pl hpssgui.vbs hpss.jar [ packaged ssm client in /tmp/ssmclientpkg.tar ] 13. 2. A cc ou nt ing HPSS m aintai ns account ing infor m ation on the usage of the sy s tem whet her the site charge s for usage or not.
An accounting policy is r equired whet her the site actuall y charges users f or HPSS usage or not . Fie ld D es crip ti on s Accounting Style. The style of accounting t hat is use d by the entire HPSS system. Valid values ar e SITE or UNIX. The defa ult value is UNIX.
manipulation operations . Account Inheri tance. A flag that i ndicates whether or not newly crea ted fil es and di rectories should automaticall y inherit the a ccount inde x used by thei r parent director y . The default value is OFF. It is onl y used if Account Vali dation has been enabled and Site- style accounti ng has be en selected.
This window allows an administrat or to view the accounti ng status and star t accounti ng. Fie ld D es crip ti on s Subsystem. The nam e of the stora ge subsystem containi ng this account ing sta tus data . Run Status. Current status of account ing run.
The fi rst type, denoted by a zero (0) in the fourth c olumn, gives the following summ a ry information about the storage used by a particul ar HPSS Account Index (AcctId) in a part icular Class Of Service (COS): • The tot al number of file a ccesses ( #Accesses) to fil es owned by the Account Index in the Class Of Service.
In the above example, line 2 shows that a user using account 634 made a total of 89 acc esses to COS 1 and has 125 files st ored in COS 1 which together total 4168147 stor age units. The s torage units reported by the report utili ty m ay be configur ed in t he Accounting Policy to repr esent by t es, kiloby tes, megaby t es, or gigaby tes.
● The sit e will need to create a local site Account Map to m aintain a lis t of the account IDs for which each user is authorized. This is a loca lly designed and written table, not supported by HPSS.
13. 2. 3. 1. 2. Sit e De fi ned A cc ou nt Ap por ti on me nt Tab le In U NIX-sty le account ing, the UID (as Account Index) maps only to a speci fic user. Some sites m ay wish to apport ion different percentages of the charges for a single UID am ong different proj ect charge codes, but without using site s ty le accounting.
13. 2. 3. 2. Acc ou nt ing In ter v al s an d Cha rge s The ti m e between accounti ng runs and the charging polic y for space usage should be developed after consulting the accounti ng requirements. The following ar e some guidelines to consi der: • Accounting should be run at a regular intervals, such as onc e per month.
Cha pte r 14 . Us er I nt er fa ce s This chapter confi guration information for the user int erfaces provided with HPSS for transfer ring fi les: • Client Application Programm ing Interface (API) • Parallel File Transfe r Protocol (FTP) or PFTP • HPSS Virtual File Sy stem (VF S) Interface 14.
write opera tions between ca che invalida tes. The default value i s 20. HPSS_API_ HOSTNAME specifi es the hostname to be used for TCP/IP listen ports cr eated b y the Client API. The defa ult value is HPSS_HOST. This value can have a signi ficant impact on data transfer performance for data t ransfers that are handled by the Client API (i.
etc/hpss. key tab. HPSS_UNIX_KEYT AB_FILE speci fies the name of t he file containing the securit y key s nece ssary for successfull y init ializing the Client API for UNIX authenti cation. The default is auth_keytab:/var/hp ss/et c/hpss.uni x.key tab.
non-HPSS Parallel FTP Daemon (DIS 2 COM PFTP Daemon). This fil e should be customized as needed. Refer to the HPSS.conf m an page or the HPSS Instal lation Guide , Appendix D for details . NOTE: it may be necessary for a site to merge older copie s of the HPSS.
# HPSS file systems rather than the local file system of the # host executing the FTP Daemon. This is highly recommended! banner <pathname/filename> # Control for logging (sent to syslog()).
• Message/re adme/bann er/shutdo wn (m essage lines) ar e text files, with the following keywords (all one characte r in l ength) recognized, identifi ed by a prece ding % : Tab le 5.
day s of the week. Ste p 4 . Cre ati ng FT P Use rs In order for an HPSS user to use FTP, a UNIX and/or Kerberos userid and password must be created. Refer to Section 3.
character device ( /dev/hpssfs0) that is used t o comm uni cate with t he HPSS VFS Daemon (hpssfsd). All POSIX I/O sy s tem c alls like open, read, writ e, cl ose, etc. are firs t handled by the VFS abstrac tion lay er after which the y are passed down to appropria te f unctions in the Kernel Module.
% make BUILD_ROOT=/tmp/vfs_client build-clnt build-fs Create a tar fi le that can be use d to build the cl ient code on the VFS machine: % cd /tmp/vfs_client % tar -cvf ../vfs_client.tar * Copy (i.e. scp) the new tar file to the c lient m achine. On the cli ent machine, untar and build t he client tree.
% rmmod hpssfs # this ensures that there isn't a pre-existing module loaded % modprobe hpssfs % make config % MAKEDEV hpssfs % /sbin/chkconfig –add hpssfs Build and install the appl ication daemon by following the instruct ions given by the build-help t arget of the makefile in the dir ectory above t his one.
% mkdir /var/hpss/cred % mkdir /var/hpss/tmp On the Core Server machine, use mkhpss to create the cli ent confi g bundle: % mkhpss Select "Create Config Bundle" to creat e a cl ient config bundle t hat contains config.
o /opt/hpss/ lib/l ibhpssuni xauth.so o /opt/hpss/ lib/l ibhpssl dapauthz.s o • If the code is installed in a non-standard loca tion (not /opt/hps s), upda te the paths in auth.conf a nd authz.conf to use the correct location. If using kerberos authent icati on, modify the /e tc/krb5.
14. 4. M ou nt ing VFS F ile sys te m s An HPSS fileset or direc tory is m a de avail able for user access by mounting it usi ng the mount( 8) comm a nd. The m ount comm and accepts the m ount input options direct ly from the command line or from the corres ponding entr y defined in the /et c/fstab file.
n o n e / d ev / s h m t m p f s d ef a u l t s 0 0 n o n e / p r o c p r o c d ef a u l t s 0 0 n o n e / sy s sys f s def a u l t s 0 0 /d e v/ V o l G r o u p 0 0 / L o g V o l 0 1 s w a p s w a p .
Mount Opti on Mount Descr iption acdirtimeo Second s to cache direc tory attribute s. The shorter this value, the more frequentl y communic ation to HPSS is require d. Synchro nization between separate vfs nodes or mount points m ay benefit f or a shorter cache period with the expense of longer latencies.
Mount Opti on Mount Descr iption maxiowidth maxiowidth princ Override for prin cipal auth Override for authorization type key Override keytab type key tab Override for keytab file stage / no stage Default is stage. This overrides the COS setting . Applic ation can override this by specifying O_NONBLOCK on the open() system call.
where <group ID> can be obta ined using the mount uti lity. The trace level can be changed while the hpssfsd is running. # echo 5 > /proc/fs/hpssfs/<group ID>/trace • Show general i n.
Cha pte r 15 . Ba ck up a nd Re cov er y This chapter dis cusses the foll owing operation s: • Backup and recover HPSS m etadat a • Backup HPSS environment • Recover HPSS user data • Handling DB2 space shortage 15.
made to the database. These logs allow DB2 to recover all changes made to the database since the time of the l ast database backup, forward to the time when DB2 was stopped by a crash, hardware failur e, power outage or whatever. It is vital that the DB2 log fil es be host ed on a highl y reliable disk systems.
“full”, “incremental” , or “delt a”. Full backups record the full cont ents of the database at a point i n time. Increm ental backups record all the changed dat a since the last full backup. Delta backups record only the data that has cha nged since the most recent backup (of an y ty p e).
15. 1. 3. O ve rv iew of th e DB 2 R eco ve ry Pro ce ss Upon discovering a dam aged DB2 contai ner, y ou must first determine the level of hardware recovery and problem deter m ination availabl e to y ou. For example, whether or not you were utilizi ng som e level of RAID and can recover a fai led disk fr om a good disk.
• {HPSS secondary metadata backup path} /subsys1 • /<metadata backup files y s tem> • Others s ite-specific files y s tems 15. 2. 2. O pe ra tin g S yst em B ac ku p It is also nec essary to per form appropriate OS backups on a regular ba sis.
prepare for and to perform the recovery process: 1. Det ermine the name of the pote ntiall y damaged volum e. Attempts to read the dam aged volume will result in Mover alar m m essages being issued t o SSM. The alarm messages will contain the name of the phy s ical volume for whic h the e rror occurred.
can be used to clean up storage resources and report what data has been lost. The recover utilit y has several options. These options a re used depending on the s everity of the damage. Refer to the f ollowing secti ons for m ore information on how and when to use these opti ons.
storage level = 1 on VV = VOL00100 path ( Fileset24: /home/bill/file2) ========= Trying to recover bitfile ========= 0786ab2c-156b-1047-8c81-02608c2f971f 00336b52 4631 10cf 00 00 00 02 storage level = 1 on VV = VOL00100 path ( Fileset24: /home/bill/file3) ========= Trying to recover bitfile ========= .
storage level = 1 on VV = VOL00100 path ( Fileset24: /home/bill/file2) lost segments from this storage level offset = 0 , length = 32768 offset = 32768, length = 32768 offset = 65536, length = 32768 At the end of the recover y , no segm ents or volumes associate d with the damaged segm ents are purged or deleted.
00226b52 4631 10cf 00 00 00 17 At the end of the clea nup, all the virt ual and physical volumes associat ed wit h the tar geted volumes will be delete d. All t he physical volum es contai ned in the virtual volume will be exported from the PVL. The media can then be imported back into HPSS to be reused.
15. 5. 1. D MS T able S pa ces Capacit y for a DMS table space is the total si ze of containe rs all ocated to the table space. When a DMS table space reache s capaci ty (depending on the usage of the table space, 90% is a possible threshold), y ou should a dd m ore spa ce to it.
HPSS Management Guide Novem ber 2009 Release 7.3 (Revision 1.0) 367.
Chapter 16. Management Tools 16. 1. U til it y Ove rv iew HPSS provides a varie ty of command- line utilit y program s to aid in the use and m anagement of the sy stem. The programs can be grouped into the foll owing major cate gories: 16. 1. 1. F ile se t a nd Jun ct io n M ana gem en t These program s work for HPSS filesets .
• dumppv_pvr - List the phy sic al volumes managed by a partic ular PVR. • lshpss - Lists various views of the system configurati on. • lsrb - Lists a list of bitfi les tha t have not be en accessed since a given date/time. • mps_reporte r - Produces human -readable output from MPS summ ary files .
• hpss_unix_user - Manages HPSS UNIX password fil e. • hpss_managetabl es – Creates, deletes and modifies databases, database tables paces, t ables, views and constrai nts. This program is normally used by mkhpss . It should be used by hand onl y by extremely knowledgeable administ rators.
can also be used to reset the number of segments count er in disk and tape storage maps. • showdis kmaps – Sends a command to the Core Server in the selec ted stor age subsystem to dump its in-mem ory disk spac e allocati on maps, t hen displa y s t hat infor m ation on standar d output.
HPSS Management Guide Novem ber 2009 Release 7.3 (Revision 1.0) 372.
App end ix A. G los sa ry of T erm s an d A cro ny ms ACI A utoma tic Media Library Clie nt Interface ACL A ccess Control L ist ACSLS A utoma ted Cartridge Sy stem L ibra ry Sof tware (Science Technol.
bitfile segme nt An internal met adata structure, not norm ally visible, used by the Core Server to map contig uous pieces of a bitfile to underlying storage. Bitfile Service Portion of the HPSS Core Server that provides a logical abstraction of bitfiles to its clients.
DEC Dig ital Equipment Cor poration. delog The process of extraction, forma tting, and ou tputting HPSS centra l log records. deregistration The process of disabling noti fication to SSM for a particular attribute chang e. descriptive nam e A hu man -readable name for an HPSS server.
fileset ID A 64-bit n umber tha t uniquely identifies a fileset. fileset name A na m e tha t uniquely i dentif ies a fileset. file system ID A 32-bit number tha t uniquely identifies an aggr egate.
IBM International Busine ss Machine s Corporation ID I dentifier IEC I nternational Electrotechnical Comm ission IEEE I nstitute of Electrical and Electronics Engi neers IETF I nternet Engineer ing Task Force Ime x Import/Export im port An ope ration in whic h a cartridge and its associated storag e space are made available to the HPSS sy stem.
local log An optional circular log mai ntained by a Log Clien t. The central log contains forma tted messag es from all enabled HPSS servers residing on the sa me n ode as the L og Client. Location Server An HPS S server that is used to help clien ts locate the appropriate Core Server and/or other HPSS server to use for a particular request.
mou nt An operation in w hic h a cartridge is either phy sically or logically made readable and/or writab le on a drive. In the case of tape cartridges, a mount operation is a phy sical operation. In the case of a fixed disk un it, a moun t is a logical operation.
physical volume An HPS S object m an aged jointly by the Core Serve r and the Physical Volu me Library that represents the portion of a virtual vo lume. A virtual vol ume m ay be composed of one or more phy sical volumes, but a phy sical volume ma y contain data from no mor e than one v irtual volum e.
request A log record message ty pe used to log some action being performed by an HPSS server on behalf of a clien t. RISC Re duced Instruction Set Co mputer/Cy cles RMS Remo vable Media Service RPC Re mote Procedure Call SCSI Sm all Com puter Sy stems Interf ace security A log record message ty pe used to log sec urity related event s (e.
stage To copy file data from a leve l in the file’ s hierarchy onto the top level in the hierarchy . start-up An HPSS S SM adminis trative operation that causes a ser ver to begin execution. status A log record message ty pe used to log processing r esults.
System Manager The Storage Sys tem Mana geme nt (SSM) server. It comm unica tes with all other HPSS componen ts requiring mon itoring or control. It also communica t es with the SSM gra phical user interface (hpssg ui) and comm and line i nterface (hpssadm).
HPSS Management Guide Novem ber 2009 Release 7.3 (Revision 1.0) 384.
App end ix B. R efe re nce s 6. 3580 Ultrium Tape Drive Se tup, Operator and Service Guide GA32- 0415-00 7. 3584 UltraScal able Tape Library Planning and Operator Guide GA32-0408-01 8. 3584 UltraScal able Tape Library SCSI Reference WB1108-00 9. AIX Perform ance Tuning Guide 10.
33. J. Steine r, C. Neuman, and J. Schill er , " Kerberos: An Authenticati on Ser vice for Open Network System s ," USENIX 1988 Winter Conference Proceedings (1988).
App en dix C. D eve lop er Ack now le dgm e nt s HPSS is a product of a gov ernment-industry coll aboration. The project approach is ba sed on the premise that no single company , governm e nt la bora.
HPSS Management Guide Novem ber 2009 Release 7.3 (Revision 1.0) 388.
An important point after buying a device IBM RELEASE 7.3 (or even before the purchase) is to read its user manual. We should do this for several simple reasons:
If you have not bought IBM RELEASE 7.3 yet, this is a good time to familiarize yourself with the basic data on the product. First of all view first pages of the manual, you can find above. You should find there the most important technical data IBM RELEASE 7.3 - thus you can check whether the hardware meets your expectations. When delving into next pages of the user manual, IBM RELEASE 7.3 you will learn all the available features of the product, as well as information on its operation. The information that you get IBM RELEASE 7.3 will certainly help you make a decision on the purchase.
If you already are a holder of IBM RELEASE 7.3, but have not read the manual yet, you should do it for the reasons described above. You will learn then if you properly used the available features, and whether you have not made any mistakes, which can shorten the lifetime IBM RELEASE 7.3.
However, one of the most important roles played by the user manual is to help in solving problems with IBM RELEASE 7.3. Almost always you will find there Troubleshooting, which are the most frequently occurring failures and malfunctions of the device IBM RELEASE 7.3 along with tips on how to solve them. Even if you fail to solve the problem, the manual will show you a further procedure – contact to the customer service center or the nearest service center
