Zimbra JVM issue with openvz

[rambeats]

Hi,
I have a 64bit Centos VPS hosted on openvz. I could install zimbra without any error. But when I try to start/stop zimbra using zmcontrol start, Getting a JVM error below,
Could not create the Java virtual machine.

mkdir : No such file or directory at /opt/zimbra/bin/zmcontrol line 35
What would be the reason? Has zimbra supports openvz?. Is this related with privvmpages size?

Please help me.

[phoenix]

[quote user="rambeats"]What would be the reason?[/QUOTE]Most likely a lack of RAM (search the forums for details)
[quote user="rambeats"]Has zimbra supports openvz?.[/QUOTE]No it doesn't, that's not a supported platform and you'll find that there are problems using it. You may get it to work, you may not.

[undersys]

You can get it to work... to an extent.

I now only have issues with starting and stopping. Some times in need to start each part manually. overall stability has been fine.
OpenVZ is crap as it can not allocate contiguous memory segments, Java expects this. Hence why ive got very small Xmx and Xmx sizes.

Unfortunately I need to work in a limited budget for my VPS.
This is for a very small deployment (under 5 users)

Here is all the settings I changed :-

USE AT YOUR RISK!


# zmlocalconfig -e mailboxd_java_options=”-server -Djava.awt.headless=true -XX:+UseConcMarkSweepGC -XX:NewRatio=2 -XX:PermSize=128m

-XX:MaxPermSize=128m -XX:SoftRefLRUPolicyMSPerMB=1 -Xms128m -Xmx128m”

# zmlocalconfig -e tomcat_java_heap_memory_percent=5

# zmlocalconfig -e zmmtaconfig_interval=3600

# zmlocalconfig -e zmstat_interval=300

# zmlocalconfig -e mailboxd_java_heap_memory_percent=5

# zmlocalconfig -e mailboxd_java_heap_new_size_percent=5

# zmlocalconfig -e ldap_common_threads=16

# zmlocalconfig -e ldap_common_toolthreads=8

# zmlocalconfig -e ldap_db_cachesize=50000

# zmlocalconfig -e ldap_db_idlcachesize=50000

# zmlocalconfig -e ldap_db_dncachesize=0

# zmlocalconfig -e zimbra_zmjava_options=”-Xms128m -Xmx128m”

# zmprov mcf zimbraMessageCacheSize 5000




Edit the following files

# nano /opt/zimbra/conf/amavisd.conf.in

Edit the folowing

“$max_servers =10″ and set it to “$max_servers =2″
Edit/add the following

“innodb_buffer_pool_size =” to “innodb_buffer_pool_size = 144245094″

“thread_cache = ” to “thread_cache = 5″

“max_connections = ” to “max_connections = 15″

[pixelplumber]

we've run 32bit zimbra on a centos 4 vps for a few years, the biggest issue we've had is that while you may allocate 50% of your system memory to the container, the JVM reports back the total system memory when it runs. So in a lot of cases it allocates far more memory to itself than the container actually supports.
undersys is right, setting the -Xms -Xmx switches sufficiently low fixed almost all of the problems we had running it. That and looking for failcnts in the user bean counters. So it can work, and pretty well too.
Having said that when we move to 7.x it will probably be 64bit in a KVM virtual machine. OpenVZ seems to have stagnated (athough they will probably pick up again after RHEL6 is released) and lxc may end up becoming the defacto container solution in another few years anyway.
Cheers.