High iowait from logger after upgrade to 4.5.9

[pi13137]

Hello all,
after upgrading from a smooth running zimbra 4.5.7 to 4.5.9_GA_1454 on my Suse 10 Linux box i've trouble with very high iowaits on my machine:

After starting zimbra iowait goes up; no idle time left. after investigation i found the zimbra logger process as source of the problem: stopping the logger everything is ok.

what are the deeper problem behind this symptoms? How can i find out? (there is no heavy traffic on this machine, so there aren't realy IO probs).

thanks for your advise.

[jholder]

Try looking in
/opt/zimbra/logger/db/(hostname).err for any mysql errors.

Also, have you run a fsck recently? How about disk space and disk format?

[pi13137]

Thanks jholder for your good tip!

The log shows me that table ...mta is marked as crashed and sould be repaired.

I try the instructions on Logger - Zimbra :: Wiki and i report if this helps.

thx!

[mrball]

I have this exact same problem - very high iowaits and 0% idle. When I shutdown logger it goes back to normal (99% idle).

However, there are no errors in the logs - just the normal mysql startup and shutdown messages

BTW - I DID have errors before, but I purged the logger db as here: Cleaning Zimbra Logger logs

Also, I noticed a script running that seems to take up all available cpu for 2-3 minutes at a time: /tmp/.swatch_script.2339

I am running 4.5.9 as well

Any ideas?

[mrball]

Originally Posted by pi13137

Thanks jholder for your good tip!

The log shows me that table ...mta is marked as crashed and sould be repaired.

I try the instructions on Logger - Zimbra :: Wiki and i report if this helps.

thx!

Does it solve your problem?

[jholder]

Originally Posted by mrball

Also, I noticed a script running that seems to take up all available cpu for 2-3 minutes at a time: /tmp/.swatch_script.2339

hmm, can you take a look at /var/log/zimbra.log? Are entries logged in there, and are there any errors?

[mrball]

Originally Posted by jholder

hmm, can you take a look at /var/log/zimbra.log? Are entries logged in there, and are there any errors?

]$ ls /var/log/zimbra.log -al
-rw-r--r-- 1 zimbra zimbra 469561902 Nov 9 02:50 /var/log/zimbra.log

It is huge. I don't see any errors... what errors should I be looking for?

[jholder]

If it's that big, then something must be wrong.

What I would do is:

su - zimbra
zmcontrol stop
cat /dev/null > /var/log/zimbra.log
zmcontrol start

and look for errors.

Do you get a lot of mail? That might explain why that file is big. . .but shouldn't be that large.

[pi13137]

@mrball:

Does it solve your problem?

probably!

call me dummy: i followed the instructions on thread http://www.zimbra.com/forums/install...working-2.html an shot up for resetting the logger and deleted as described in the thread /opt/zimbra/logger/db/data/*.
so i now set up a completly new/fresh zimbra.

thanks god this was not a production-system

[mrball]

Originally Posted by jholder

If it's that big, then something must be wrong.

What I would do is:

su - zimbra
zmcontrol stop
cat /dev/null > /var/log/zimbra.log
zmcontrol start

and look for errors.

Do you get a lot of mail? That might explain why that file is big. . .but shouldn't be that large.

No errors...

It's big because logrotate was not working on that machine but that has now been rectified. It does seem to log quite a bit though - postfix emails received, amavis scans, etc.

However, the high iowait problems are persisting. any ideas? And zmlogger seems to restart itself every once in a while even after I stop it

I have resetted the db. I have even tried moving the whole /opt/zimbra to a partition on a different hdd and rsyncing the files across. fscks are all OK.

Any ideas?