[SOLVED] Very high CPU usage with 7.1.1

[MichaelB12]

Hi,

From Zimbra version: 7.1.1_GA_3196 under CentOS 5.5 x64, my CPU still very high all the time if i compare with previous version. That overload the CPU of my main VM server.

Release 7.1.1_GA_3196.RHEL5_64_20110527011124 CentOS5_64 FOSS edition

I suppose is a bug.. ? Can i downgrade to 7.1.0 without problems ?

When the next release 7.1.2 will be released ?


See attach... (CPU of the Zimbra VM)


Thanks in advance,
Michael

[mtamas]

Hello Michael,

I have tried to upgrade my test system.
I have verified it. My CPU is running on 70% - 90% all the time.

About the downgrade:
"Downgrading to version 7.1.0 from 7.1.1 is not supported."

Cheers,
Tom

[rotorboy]

I've had the opposite. My loads were at 40+% all the time until I upgraded. Now it's dropped to 3-8%.

[Labsy]

What does "top" say, which processes are most CPU hungry?
Did upgrade pass without warnings and errors?

[MichaelB12]

Yes upgrade pass without problem like each times for me... That process below run every ~2 minutes and take almost 100% of CPU for ~1 minute.. I already had something similar in a older version may be one or two years ago but was fixed..


See which process take CPU:

zimbra 13156 86.6 8.2 613216 169004 ? Sl 10:42 0:12 /opt/zimbra/java/bin/java -XX:ErrorFile=/opt/zimbra/log -client -Xmx256m -Dzimbra.home=/opt/zimbra -Djava.library.path=/opt/zimbra/lib -Djava.ext.dirs=/opt/zimbra/java/jre/lib/ext:/opt/zimbra/lib/jars:/opt/zimbra/lib/ext-common:/opt/zimbra/lib/ext/clamscanner -Dpython.cachedir=/opt/zimbra/data/tmp/jython-cachedir org.python.util.jython /opt/zimbra/libexec/zmconfigd


I discover that bug.. may be is my problem.. ?
Bug 60009 – sys cpu spike caused by jetty


Thanks for your answers,
Michael

[FritzBrause]

The high server load value together with 500 MByte swapped suggest that you have a memory problem. The high CPU usage may result from disk IO reading and writing memory pages to the swap partition.

Tune the following parameters:

/opt/zimbra/conf/my.cnf

innodb_buffer_pool_size        = 384M

/opt/zimbra/conf/amavisd.conf.in

$max_servers =  2

then as zimbra user do

zmlocalconfig -e mailboxd_java_heap_memory_percent 20

Also you might adjust your Systems Swapping behaviour:

#
# Tuning Kernel VM Settings
#

sysctl -w vm.swappiness=0
sysctl -w vm.vfs_cache_pressure=50
sysctl -w vm.min_free_kbytes=16384
sysctl -w vm.dirty_ratio=10
sysctl -w vm.dirty_background_ratio=5


/etc/sysctl.conf

#
# Prevent Kernel from swapping and too much IO
#

vm.min_free_kbytes=16384
vm.swappiness=0
vm.vfs_cache_pressure=50
vm.min_free_kbytes=16384
vm.dirty_ratio=10
vm.dirty_background_ratio = 5

Restart server.

Regards Andreas

[MichaelB12]

Yes i saw that some swap is used, it appear after some weeks of usage.. But i don't think is the cause of this problem. If i restart my server no swap is used and cpu spike continue.. But thanks for your tweaks i will test it later on my server.

I saw that someone else have a similar problem in this thread:

Zimbra 7.1.1: High CPU spikes every 5 minutes


Michael

[halley]

Hi,
any information for fix this?
thanks

[MichaelB12]

In my case, i fixed the problem...
It exist a patch for 7.1.1 GA (Zimbra Collaboration Suite Open Source Edition)

Open Source Edition Downloads: Enterprise Messaging and Collaboration Software by Zimbra

My CPU back to a normal value. See my attach...


Michael

[PhD]

thats interesting - the release notes didnt mention anything about it either...
it would be good to have this addressed for NE users.