Page 1 of 1

Performance

Posted: Tue Dec 13, 2016 6:07 pm
by zgokan
Hello

I have 190 mailboxes in the server. The RAM is 16G. Cpu 8 is the core. It's not enough.
The server sometimes freezes. It gets better when you restart it. (note: My hard drives are SATA.)

What do I have to do to get the problem out?

Code: Select all

[zimbra@mail root]$ zmcontrol -v
Release 8.7.0_GA_1659.RHEL7_64_20160628202714 RHEL7_64 FOSS edition.


Code: Select all

[root@mail ~]# free -gh
              total        used        free      shared  buff/cache   available
Mem:            15G        5,2G        456M        320M        9,9G        9,7G
Swap:            9G          0B          9G

Re: Performance

Posted: Tue Dec 13, 2016 9:58 pm
by JDunphy
Would need a bit more information to help with a guess?

Have you grabbed any other stats? Do you have an educated guess of which subsystem? ... disk, cpu, interrupts, memory, network, etc.
Dedicated hardware or Virtualized (KVM, Xen, vmware, etc)? What type of raid configuration are you running with those SATA disks?
What is your network connected speed by your host and have you done anything to tune it if connections are dropping or being denied... ie. https://access.redhat.com/sites/default/files/attachments/20150325_network_performance_tuning.pdf

Code: Select all

vmstat 5
iostat 5
ps axo pid,ppid,rss,vsz,nlwp,cmd
cat /proc/interrupts
cat /proc/cpuinfo
grep -i error /opt/zimbra/log/*.log
tail -f /opt/zimbra/log/myslow.log


What does freeze mean? The users are feeling lockups/delays or you are rebooting the server because it has locked up and unresponsive from the network and your out of band console access hangs.
What is the mix of web users to pop/imap/etc. How many established connections when it happens? dmesg?, Anything in /var/log/messages? , etc, etc.

I tend to run vmstat and drill deeper from there but you need to exhaust looking through your logs for errors or warnings. Could be as easy as disk errors but the logs would have that information. Depending on how the hardware fails it can lockup the bus until a kernel watch dog fires and "unfreezes" everything. It's fairly obvious with dmesg if that is happening and easy to verify from there.

I tend to focus on kernel subsystems and work back from there. Lots of ways to isolate the problem.

HTH