I'm facing the same problem right now, I'll try to give as much valuable informations as I can.
This is a fresh (1 month) installed server.
* OS: CentOS release 5.5 (Final)
* Zimbra: Release 6.0.8_GA_2661.RHEL5_20100820021155 CentOS5 NETWORK edition.
* no iptables installed
Every few minutes the Nagios server goes wild and warns me that some Zimbra services are down.
- Tests with telnet from the client to any zimbra port confirm the issue.
- The problem lasts for like 1 minute and then goes back to normal.
- Only the client suffers from non responsive server.
- Other hosts in the same LAN can access without any problem (unless I try to connect too often).
tcpdump on zimbra server of the problem (port 110)
This proves that Zimbra is receiving the packets but does not transfer them to the service because nothing changes in file /opt/zimbra/log/mailbox.log
[root@zimbra log]# tcpdump port 110
tcpdump: verbose output suppressed, use -v or -vv for full protocol decode
listening on eth0, link-type EN10MB (Ethernet), capture size 96 bytes
11:48:17.775082 IP nagios.domain.tld.58137 > zimbra.domain.tld.pop3: S 2582162921:2582162921(0) win 5840 <mss 1380,sackOK,timestamp 661513095 0,nop,wscale 6>
11:48:20.771473 IP nagios.domain.tld.58137 > zimbra.domain.tld.pop3: S 2582162921:2582162921(0) win 5840 <mss 1380,sackOK,timestamp 661513845 0,nop,wscale 6>
11:48:26.770916 IP nagios.domain.tld.58137 > zimbra.domain.tld.pop3: S 2582162921:2582162921(0) win 5840 <mss 1380,sackOK,timestamp 661515345 0,nop,wscale 6>
Don't know what's going on here.
I suspect Zimbra to block access from clients if too many connections are made, (or maybe too many opened connection at the same time?)
Some advice is needed. I would like to monitor my server without being waken up at night for nothing :P