Unusual memory and swap usage for Zimbra on CentOS4 Linux - Help please
We're using Zimbra 3.1 on CentOS4 Linux with 1GB RAM. The Zimbra client is quite slow to respond to requests, so I did a quick check of top and got the following unusual stats. High load average and unusual amount of swap being used.
Does anyone know what would cause only 470MB of RAM to be used out of a possible 1GB available RAM and then 512MB of swap? Very weird.
This server is currently supporting 15 users.
Any suggestions appreciated.
[admin@mail ~]$ top
top - 18:47:29 up 1 day, 3:58, 1 user, load average: 4.99, 3.14, 2.61
Tasks: 108 total, 3 running, 105 sleeping, 0 stopped, 0 zombie
Cpu0 : 2.5% us, 13.2% sy, 0.0% ni, 49.1% id, 35.2% wa, 0.1% hi, 0.0% si
Mem: 1043912k total, 469896k used, 574016k free, 992k buffers
Swap: 1052248k total, 513748k used, 538500k free, 37292k cached
No we did that as part of the install process
Thanks for the reply, we actually did that bug fix as part of the install process to get it up and running.
Originally Posted by phoenix
1. This server is running within VMWare server as a VMWare image.
2. CentOS 4.4 OS
3. The server load average is frequently between 1 and 5 with a high amount of swap usage, even though there is plenty of available RAM. This impacts most noticably on the Web UI causing the "server slow warning" to be displayed to users.
This problem has occurred after running the system in production for a week. It did not show up during the install/testing process. It's totally bizarre.
Here's another top snapshot.
[admin@mail ~]$ top
top - 11:15:24 up 1 day, 20:26, 1 user, load average: 1.14, 1.63, 2.18
Tasks: 99 total, 1 running, 98 sleeping, 0 stopped, 0 zombie
Cpu(s): 0.0% us, 1.1% sy, 0.0% ni, 94.7% id, 4.2% wa, 0.0% hi, 0.0% si
Mem: 1043912k total, 393024k used, 650888k free, 652k buffers
Swap: 1052248k total, 487688k used, 564560k free, 35216k cached
Solution: Poor performance on CentOS 4.3
Found the solution to this problem. Turns out that it was a CentOS 4.3 kernel problem (or bug).
The latest Kernel update (2.6.9-42.0.2.EL) fixed the problem. Now the SAME server does not use any swap at all and is responding much quicker.
So there appears to be a problem with CentOS 4.3 and the default kernel, 2.6.9-34.0.2.EL which causes performance and swap issues. This may affect just Java applications or Java/VMWare guest OS - who knows.
Just very glad that it's been fixed.
So for anyone else running CentOS 4.3 and experiencing terrible performance and high swap usage, I suggest you run update and get the latest kernel.