UNSOLVED
Unable to schedule hfscheck level ?
Hi,
Hfs check is consuming more then 8 hours in a 4 tb multinode environment and we want to run refcheck in weekdays and full on weekend. I copy hfsscheduleconf.xml and edited the copy as below.
admin@ave:/usr/local/avamar/var/>: more hfsscheduleconf.xml.orig
then I run below command to test it.
hfscheck_cron --useschedule --scheduleconffile=/usr/local/avamar/var/hfsscheduleconf.xml.orig
But saw that it is forcing to full as you can see at below.
2011/10/10-11:11:04 ============================== Running hfscheck_cron ==============================
2011/10/10-11:11:04 ps -efww
2011/10/10-11:11:04 avmaint nodelist --short --xmlperline=500
2011/10/10-11:11:06 ======== starting hfscheck ========
2011/10/10-11:11:06 using activityfile: /usr/local/avamar/var/hfsscheduleconf.xml.orig
2011/10/10-11:11:06 performing refcheck hfscheck
2011/10/10-11:11:06 avmaint hfscheck --full --concurrentindexstripes=2 --keepmin --modified=2 --throttlelevel=1 --avamaronly --xmlperline=99
2011/10/10-11:12:11
Is there any ideas for scheduling hfs check level ?
Thanks
Avamar Exorcist
462 Posts
896
1
Posted October 11th, 2011 06:00
It looks like the main perceived issue here is related to HFScheck performance.
First of all, since this is Avamar v5, we should stick to rolling checkpoints only. A full checkpoint sometimes comes in useful but should not be necessary.
Many factors contribute to HFScheck performance including GSAN capacity, overall change rate, node type and disk configuration and scheduling but we would expect a typical rolling HFScheck to complete within 5 hours.
Some practical advice would be to avoid scheduling backups and replication during HFScheck and you will find the HFScheck completes more quickly, (leaving time for a longer backup window).
You can easily check if HFScheck is overlapping with any other activities by downloading and running the latest copy of the sched.sh script. The link is below.. see also esg110685
ftp://avamar_ftp:anonymous@ftp.avamar.com/software/scripts/sched.sh
The same script also has a --hfscheck option which will output performance figures for recent HFScheck operations. You can use this to compare their performance before and after making scheduling changes.
I would also recommend opening an SR with support to request that the system be updated with lastest GSAN as there have been some improvements (including for HFScheck performance) in later releases of v5.x
Hi Nicolas,
Thanks for your reply.As far as I understand it is not possible to change hfs check schedule. Actually our backup window`s 1-2 hours conflicts with hfs check. We have nearly 80000 stripes and we have nearly 600 virtual clients. We are running 300 vm backup per day and cbt has been installed nearly all clients except 10-15 but Avamar can backup nearly 150-180 per day in 12 hours. By the way rolling hfs check validates nearly 80000 stripes every day.
Actually we opened a support ticket about hfs check duration and engineer has been check the env, and report that 8-10 hours duration may possible for this environment,
Avamar Exorcist
462 Posts
895
0
Posted October 12th, 2011 04:00
Hi Can, If the system is highly loaded then for sure a longer HFScheck would be expected.
Another option to speed things up would be to increase the number of storage nodes in the grid. This will effectively reduce the amount of work per node when performing HFScheck, increase backup sessions. Provided it doesn't already have the maximum 16 data nodes, that is..
Nayak2010
50 Posts
896
0
Posted October 10th, 2011 12:00
What is the version of avamar ?