We have a community edition zcs-3.1.0_GA_332.RHEL4 running on Red Hat Enterprise Linux ES release 4 (Nahant Update 3), on an HP DL120 with 1x Intel Xeon 2.8GHz, 1GB RAM and 2x 250GB SATA drives, which are doing software RAID mirroring.
The system does approx 80 users with very average amounts of mail, nothing like what some people have posted their systems are doing (no 10 000 messages every half hour or anything like that). All Zimbra services are switched on, although switching off the anti-spam etc. did not solve our problem.
The problem is that the load average sits between 7 and 11, all the time. The java and amavis processes are always near the top of "top", and after not very long the system starts falling apart. Mail stops delivering, giving a few error messages.
This is one:
May 24 12:05:01 zimbra postfix/qmgr: 5389BEB0713: to=<firstname.lastname@example.org>, relay=none, delay=2119, status=deferred (delivery temporarily suspended: connect to zimbra.domain.org.za[196.xx.xxx.x]: read timeout)
This is another:
May 24 11:07:45 zimbra postfix/qmgr: 9695DEB0676: to=<email@example.com>, relay=none, delay=-2892, status=deferred (delivery temporarily suspended: connect to 127.0.0.1[127.0.0.1]: server dropped connection without sending the initial SMTP greeting)
May 24 11:07:18 zimbra postfix/smtp: 0268EEB061A: to=<firstname.lastname@example.org>, relay=127.0.0.1[127.0.0.1], delay=-2315, status=deferred (lost connection with 127.0.0.1[127.0.0.1] while sending end of data -- message may be sent more than once)
We also had another one regarding not being able to connect to Clam on port 3310, despite nobody touching Clam.
The machine also runs Squid and Samba, and so we had read in a previous post that changing the following would help:
[zimbra@zimbra ~]$ zmlocalconfig | grep mem
mysql_memory_percent = 30
tomcat_java_heap_memory_percent = 20
Is there something we're doing wrong? Is the software RAID mirroring a bad idea? Has anybody else had the same issue?
At the office we run a dodgy small HP with 1GB RAM, no mirroring on CentOS, and we don't have any of the above issues.