Question & Answer
Question
During EXPORT SERVER processing, very poor throughput rates to the target Tivoli Storage Manager/IBM Spectrum Protect server might be observed. To be exposed to the symptom documented here, your server must be configured to have an archive copygroup configured with RETINIT=EVENT and the usage of the SET EVENT command on the client. This technote documents how to verify if the source server is prepared for the export and what to do if not.
Cause
In the sample export, the following instrumentation statistics were reported on the source server for the export thread:
Thread 117 XiExport parent=116 22:46:34.762-->23:07:23.287
Operation Count Tottime Avgtime Mintime Maxtime InstTput Total KB
----------------------------------------------------------------------------
Data Copy 3493 0.008 0.0000 0.000 0.000
Network Recv 3 32.731 10.911 9.464 12.274 0.0 0
Network Send 2388 3.097 0.001 0.000 1.391 24647.8 76351
Acquire Latch 89 0.001 0.000 0.000 0.000
Acquire XLatch 6106 0.006 0.000 0.000 0.003
Thread Wait 11873 7.935 0.001 0.000 0.107
Unknown 1204.744
----------------------------------------------------------------------------
Total 1248.525 61.2 76351
Note that the vast majority of time is spent in the unknown section.
A thorough review of the source server in this case confirmed the bottleneck was with the database of the export source server, which had a huge amount of database page latches to be processed.
If you are exporting a V5 based server, consult the output of the
Q COPYG * * * type=archive f=d
command.
If you find a copygroup that is configured for Event initiated retention, you can use the SHOW IMVAR command and verify if retInitGood is "True":
tsm: DR550>SHOW IMVAR
Inventory Global Variables:
Last Object Id : 0 169966373
HWM Object Id : 0 169966593
Last Name Id : 0 0
HWM Name Id : 0 1
HWM Compression Object Id : 0 1
Expiration Interval : 0 Hours
Sort Delete Active : False
Expiration Running : False
Compression State Known : True
SET OBJECTR Begin Time : 12/00/0/, 00:00:00
SET OBJECTR End Time : 12/00/0/, 00:00:00
SET OBJECTR lower ObjId : 0.0
SET OBJECTR upper ObjId : 0.0
retInitGood : False
If, as in the above example, retInitGood is False, run the following command to prepare the server for optimized throughput for the export job:
- Issue the TSM console (dsmadmc) command:
REPAIR EXPIRATION * TYPE=ARCHIVE FIX=YES EVENTBASEUSED=FIX
- This command only works if retInitGood is False. The asterisk "*" stands for the client node list that is supposed to be processed. It means all nodes. This ensures that all entries in Expiring.Objects table are being processed and indexed.
- After successful completion of REPAIR EXPIRATION, stop and restart the TSM server to activate the new optimized lookup algorithm.
- Use the command SHOW IMVAR again to verify that retInitGood is True.
To be exposed to this symptom, the source server must have been installed at a server version prior to 5.3.6 or 5.4.2.
Was this topic helpful?
Document Information
Modified date:
17 June 2018
UID
swg21976336