MFE SCF: Recommended Method for TRU RECLAIM

SECTION 1 contains the recommended procedure to implement TRU SCAN and RECLAIM as a manual process controlled by TRU commands. This is the recommended method which is used at most sites.

SECTION 2 contains a list of items to consider when designing the optimal RECLAIM solution for your site. 

SECTION 1: Recommended procedure to setup and run TRU RECLAIM

1. Query and record your existing SRP Pool utilization before and after RECLAIM activity.

2. Add these TRU parameters to your SCF Initialization member:

***** TRU DEFINITION START ***** 
*
SCF.TRU.ENABLE=NO                
SCF.TRU.DEBUG=NO  
SCF.TRU.OFFLINE=NOPROCESS 
SCF.TRU.RECLAIM.METHOD=1  
SCF.TRU.SCAN.TASK.LIMIT=5  
SCF.TRU.RECLAIM.TASK.LIMIT=5  
SCF.TRU.SCAN.STCNAME=EMCTRU  
SCF.TRU.RECLAIM.STCNAME=EMCTRU                            
SCF.TRU.RECLAIM.DSPREFIX=dsn-prefix
*                                    
SCF.TRU.DEV.INCLUDE.LIST=ccuu-ccuu                    
*SCF.TRU.DEV.INCLUDE.LIST=ccuu-ccuu           
*SCF.TRU.DEV.EXCLUDE.LIST=ccuu   
*
SCF.TRU.SCRATCH.POST.TYPE=DEVICESIZE           
SCF.TRU.SCRATCH.POST.PCT=100  
SCF.TRU.SCRATCH.POST.MIN=9999999 
SCF.TRU.SCRATCH.POST.MAX=9999999  
SCF.TRU.SCRATCH.RECLAIM=0   
SCF.TRU.RECLAIM.POST.TYPE=DEVICESIZE  
SCF.TRU.RECLAIM.POST.PCT=100 
SCF.TRU.RECLAIM.POST.MIN=9999999 
SCF.TRU.RECLAIM.POST.MAX=9999999  
SCF.TRU.RECLAIM.SCRATCH.WAIT=100  
SCF.TRU.RECLAIM.SYSVTOC.HOLDLIMIT=1000
SCF.TRU.RECLAIM.SYSVTOC.WAIT=200
SCF.TRU.RECLAIM.STRESS.MONITOR=YES                       
SCF.TRU.RECLAIM.STRESS.WAIT=15      
*          
***** TRU DEFINITION END ***** 

3. Create the new started task EMCTRU that is used for both SCANs and RECLAIMs. Use this JCL:

//EMCTRU  PROC CCUU=,TYPE=RECLAIM              
//ESFTRURC EXEC PGM=ESFTRURC,PARM='&TYPE,&CCUU'
//STEPLIB DD DISP=SHR,DSN=your.linklib   
//SCF$nnnn DD DUMMY                             
//SYSABEND DD SYSOUT=*                         
//ESFTRURC DD SYSOUT=*

4. Activate the new TRU parameters. 

* If you start or recycle SCF, the TRU parameters are automatically activated.

* If you update TRU parameters for an active SCF, you must enter these two commands:

F emcscf,INI,REFRESH                 - This command updates the SCF INI parms

F emcscf,TRU,REFRESH               - This command updates the TRU parms

F emcscf,TRU,START,ccuu-ccuu

5. You should now display the active TRU environment information.

* To get details on the active TRU environment and parameter settings enter command: 

F emcscf,TRU,DISPLAY

F emcscf,TRU,DEVICE,ccuu

6. When you are ready to perform the SCAN and RECLAIM processes, issue these SCF commands: 

F emcscf,TRU,ENABLE

F emcscf,TRU,SCAN,ccuu-ccuu 
* Note: Wait 15 minutes before submitting RECLAIM *

F emcscf,TRU,RECLAIM,ccuu-ccuu 
* Note: RECLAIM can take a long time.  When the RECLAIM activity is finished perform the TRU,DISABLE *

F emcscf,TRU,DISABLE

These TRU commands can be done in a batch job or with operator commands. These can be submitted with an automation or scheduling system.

If you want to RECLAIM during a specific time window, issue the TRU,ENABLE and TRU,SCAN and TRU,RECLAIM commands at the beginning of the time window, and then the TRU DISABLE command at the end of the time window. 

7. In order to verify TRU processing is working properly and is optimally configured, these files should be reviewed:

• SRP Pool query output (before and after RECLAIM)
• SCF STC output
• SYSLOG that includes the entire TRU processing
• EMCTRU STC outputs (TRU SCAN and RECLAIM)

SECTION 2: Things to Consider when setting up your TRU RECLAIM environment.

1. Multi-LPAR considerations:

• How many LPARs are in your configuration?
• How many LPARs do you want to run RECLAIM on?
• Be careful not to run the same device RECLAIM on multiple LPARs. This can result in wasted processing and potential contention issues. The best way to accomplish this is to specify unique TRU device ranges on each LPAR running TRU. Only devices available to this LPAR are specified on the SCF parameter SCF.TRU.DEV.INCLUDE.LIST and on the SCAN or RECLAIM commands issued on that LPAR.

2. SRP considerations:

• What is your current SRP % allocated?
• Is the SRP% stable or is it going up?
• IMPORTANT: It is recommended that you track your SRP utilization on a weekly (or daily) basis in order to observe trends or issues. At the end of this article, there is a batch job that you can use to query your SRP pool utilization.

3. Device considerations:

• How many volumes do you want to RECLAIM? This is important to know in order to estimate the time required to perform all RECLAIMs.
• Are any of these volumes EAVs? If so, how many EAV volumes are you RECLAIMing and what size are they? 

• Are any of the volumes managed by SMS? If the volume is initialized for SMS, it must be part of an SMSPLEX on the LPAR where TRU is running.
• These volumes should not be RECLAIMed:
  • Volumes that are OFFLINE should not be RECLAIMed.
  • Volumes for high-speed online applications should not be RECLAIMed when performance is critical.
  • Volumes that have constantly changing data can cause contention and are not good candidates for RECLAIM. This includes volumes for SORT, TEMP, PAGE, SPOOL, JES Checkpoint, Coupling Facility, SYSRES, database reorgs, and volumes being compressed or defragmented.

4. Workload Contention considerations:

1. • If you run rRECLAIMs during a particular time window, be sure to consider any other workloads that may be active simultaneously. Workloads that have high I/O activity on volumes that are being RECLAIMed can result in I/O contention and longer reclaim job execution times. For example, large DB2 database systems and reorgs, data compressors, and other internal applications can result in contention with the RECLAIM process.
   • The XCF address space (XCFAS) is a service provider for applications and subsystems that communicate and serialize with other members in a sysplex. Depending on the XCF processing occurring, an increase in XCFAS CPU usage and I/O activity can occur. If this happens, it is possible that active TRU RECLAIMs could be affected.
   • Performance collection products that collect hardware statistics for the coupling facilities can also cause contention.

Sample Batch Jobs to Query SRP and Pool utilization

Run both jobs:

1. Run this job to obtain your SRP and pool usage information:

//Job1 jobcard-----
//SNAP1    EXEC PGM=ESFGPMBT,REGION=4M
//STEPLIB  DD   DISP=SHR,DSN=your.linklib
//SCF$nnnn DD DUMMY
//SYSPRINT DD   SYSOUT=*
//GPMPRINT DD   SYSOUT=*
//SYSABEND DD   SYSOUT=*
//SYSOUT   DD   SYSOUT=*
//QCOUTPUT DD   SYSOUT=*
//GPMINPUT DD   *
    QUERY SRP       LOCAL(UNIT(ccuu))
    QUERY POOLS     LOCAL(UNIT(ccuu))
/*

2. To obtain poolname details, use these GPMINPUT statements in the above job.

The poolnames are found in the output of step 1.

//GPMINPUT DD   *
    DISPLAY(       -
           LOCAL(UNIT(ccuu))
           DISPLAY(                 -
               LOCAL(UNIT(ccuu))    -
               TYPE(THINPOOL) POOL(poolname))
    QUERY THINDEV LOCAL(UNIT(ccuu)) POOL(poolname)
    QUERY DATADEV LOCAL(UNIT(ccuu)) POOL(poolname)
/*

3. To show all the individual LOCAL and REMOTE device allocations, use these GPMINPUT commands in the above job.

//GPMINPUT DD   *
    QUERY ALLOC (UNIT(ccuu))
    QUERY ALLOC RMT(UNIT(ccuu) PATH(nn))
/*