ZCS 6.0.5 - high cpu usage, freeze machine

[stich86]

hi guys,

after 2 week i've got another crash of ZCS server. I'm running it on a virtual macchine (2 CPU, 4GB ram, iSCSI storage) on Citrix XenServer 5.5

Today, when I enter on the admin console to add a new user, after click on the button the interface freeze, and on the console i've full memory and all two cpu at 100%. I cannot ping or ssh into the machine, only method to restore is a force reboot

I've also noticied that after some days logger and stats stop to working, and no zmloggctl restart/zmstatctl restart works..

There are some logs to check why this happening? Installation is on SLES10 SP3 64bit and filesystem are reiserfs

Thanks!

[tonster]

Do you have hyperthreading enabled on the host machine? We've seen a number of issues with hyperthreading being enabled. Beyond that, I'd look to see what was logged in mailbox.log and zimbra.log, as well as what was happening on the host machine when the system started going bad.

[stich86]

still have this issue also after 6.0.7 upgrade

I've left a "top" on the console of VM to see what's up when the machine hangs. But the top doesn't reporty nothing of special..

machine had a load average of 2.68, cpu at 78%sys, after this TOP freeze (same time of syslog stop) and stay with 100% CPU until a force reboot.

This is the last entry on /var/log/messages before hang:

Sep 17 16:19:00 zm01 zimbramon[17703]: :::47A646A6-C277-11DF-823C-D938DCBACED5:::,0,0,0,,2588607,0,0,0,0,6,5,11,1,765941
Sep 17 16:19:03 zm01 zimbramon[17707]: 17707:info: zmstat mtaqueue.csv: timestamp, KBytes, requests:: 09/17/2010 16:19:03, 0, 0
Sep 17 16:19:04 zm01 zimbramon[17618]: 17618:info: zmstat vm.csv: timestamp, r, b, swpd, free, buff, cache, si, so, bi, bo, in, cs, us, sy, id, wa, st, MemTotal, MemFree, Buffers, Cached, SwapCached, Active, Inactive, HighTotal, HighFree, LowTotal, LowFree, SwapTotal, SwapFree, Dirty, Writeback, AnonPages, Mapped, Slab, CommitLimit, Committed_AS, PageTables, VmallocTotal, VmallocUsed, VmallocChunk, loadavg:: 09/17/2010 16:19:04, 9, 1, 1052180, 24196, 65880, 607896, 3, 1387, 1035, 14068, 921, 462, 56, 10, 27, 7, 0, 4194476, 24196, 65880, 607896, 116104, 2993324, 985320, 0, 0, 4194476, 24196, 1052248, 68, 187284, 0, 3238420, 33264, 67612, 3149484, 6675236, 17132, 34359738367, 9060, 34359729299, 2.03

we have other machines that run intense java application like zimbra, and the "only" difference between those and this is the kernel version:

on zimbra is: 2.6.16.60-0.58.1-xen

on other machines are: 2.6.16.60-0.54.5-xen

I've read same problem on CentOS cause by a kernel bug, but nothin regarding SLES

Can anyone post its kernel version of machine that runs zimbra?

I want to run YOU update (that contains a lot of security fix included kernel) but I don't know if this can creare other problems (like depedency on new RPM?)
Thank you

[tonster]

It sounds like you'd likely find benefit from a YOU update. We do test against the latest patch versions, and recommend that customers stay updated so you're not going to run into dependancy issues unless you're, for example, running ZCS SLES10 binaries on SLES11. Your issue really sounds like an OS or kernel issue rather than a ZCS one. It's extremely rare for ZCS to get itself into a state in which it actually locks up the system.

[stich86]

so I can update without problems? The new rpm doesn't affect the current zimbra installation right?

I see you have Zimbra on SLES10, what kernel version do you have?

thank you!