We have a multi server setup with two ldaps, four proxy/mta servers and four mailbox servers. We have hardly any users, yet.
Both LDAP servers seem to need a lot of CPU power, which is not to be expected. It seems that a certain Java process is started at a regular interval. This java process turned out to be a set of calls to zmprov, consuming a lot of CPU power.
Almost every day the slave LDAP hangs. The server is still alive, connect to port 389 is possible but returns no results. If we're lucky we can get to the console and try to login, but the server turns out to be too occupied to actually let us in. The only remedy is to powercycle the server. Everything in the platform hangs or responds extremely slow until either the LDAP slave dies completely or restarts.
OS: Ubuntu 8.04 LTS 64bit on VMware ESX 4.0
Zimbra: 6.0.3 NE
The master keeps running.
The is nothing unusual in the zimbra logs. The only strange thing I found in the kernel log was:
Dec 10 07:20:06 ldapslave kernel: [87153.172904] vmstat trap divide error rip:40222f rsp:7fffbcf3ca70 error:0
Any help would be appreciated!