10-30-2014 04:15 AM
Hi all,
Our NBU 7.6.3 deployment is still in pre-production and yesterday, for the first time we ran a couple of test VM backups. The backup completed successfully however almost immediately after the backup completed, and still 24 hours later, I see heavy disk reads on the LUN that hosts the Dedupe store. If I look at the Disk tab within Windows Resource Monitor I see the following (see attachement).
This storage is connected via iSCSI if that makes a difference...
Is this expected?
Thanks in advance,
Graham
Solved! Go to Solution.
11-14-2014 03:54 AM
Hi all, sorry for the lack of updates. As above the problem went away when we recreated the MSDP volume on a thick provisioned LUN. Thanks for your help.
10-30-2014 04:18 AM
Another attachment showing NIC bandwidth history on the storage unit.
10-30-2014 04:55 AM
I'd say it is the normal maintenance jobs in your MSDP disk pool.
Have a look in the file Q:\dedupe\log\spoold\storaged.log
It should give an idea of what it is currently doing.
The spoold process should also be quite busy.
/A
10-30-2014 05:12 AM
Thanks for your help. The log is ful of these:
October 30 01:11:10 ERR [000000000A01AF40]: 25048: _storeGetCRCCheckCandidateFromJournal: can't fsync file Q:\Dedupe\databases\datacheck\crc_check_write: (invalid file descriptor)
10-30-2014 06:58 AM
I have never really seen that error, but the fsync would indicate it is unable to flush data to disk. Perhaps you should open a support case with Symantec.
In the meantime: Do you use antivirus scanner? Perhaps try disabling scan of Q drive, or relevant processes such as spoold.exe
/A
10-31-2014 02:53 AM
Likely - MSDP pool does a lot of housekeeping (space reclaim, integrity checking etc) when idle.
http://www.symantec.com/docs/HOWTO89134
http://www.symantec.com/docs/TECH124914
10-31-2014 09:20 AM
I know nothing of this facet but if I were to see messages along the lines of "datacheck" and "crc " I'd be thinking: data corruption. Worrying...
JIm
10-31-2014 02:34 PM
I've opned a case with Support. It seems to partly be down to the fact that the MSDP LUN was thin provisioned. I've since recreated the LUN as think provisioned and am using nbperfchk to make sure the storage is up to the task. I'll update this thread when I know more.
11-14-2014 03:54 AM
Hi all, sorry for the lack of updates. As above the problem went away when we recreated the MSDP volume on a thick provisioned LUN. Thanks for your help.