Just a quick followup: After I added a second 512MB swap file on January 12th I haven't had an invoked-oom failure. I have had some high swap activity and will consider deleting the 512MB swap file and creating a 1024MB swap file. A search of an older /var/log/messages file shows that several processes were being killed by the oom manager, not just mysql. As performance isn't an issue I will change some of the mysql settings to reduce its memory footprint.

For a low volume site this basic linode server looks to be adequate to run my ClearOS 7 Home Essentials with Zarafa. If I needed more performance in a real work environment I would opt for the $20.00/month 2 core, 4GB RAM option.

Thanks Tony,

I have created another 512M swap file and set it at a lower priority. I will track usage and check for invoked-oom error messages. Here is a new free -m and cat /proc/swaps.

[root@nss init.d]# free -m
total used free shared buff/cache available
Mem: 1997 1693 33 45 270 200
Swap: 1023 446 577

[root@nss init.d]# cat /proc/swaps
Filename Type Size Used Priority
/dev/sdb partition 524284 457692 -1
/swapfile file 524284 0 -2

For Zarafa gurus out there, I would like to disable starting zarafa-ical, monitor and search if someone knows where they are started. I will try removing the scripts from /etc/rc.d/init.d to see if that works. Turns out that stopping the webapp web server also disables activesync so that needs to stay up. I wil research ways to minimize mysql memory usage (if that's possible).

This only seems to happen when I add one of the email accounts to a new computer and it starts downloading thousands of emails and attachments. Shouldn't need to much of that going forward.

I will track this for the next few days and report if this resolved the invoked-oom errors.

Here is the free command on my server:

[root@nss ~]# free -m
total used free shared buff/cache available
Mem: 1997 1598 37 7 361 335
Swap: 511 48 463

Yes, I am getting occasional invoked-oom messages in my messages log for mysql that correspond to the failures. It looks like mysql gets restarted but I suspect that one or more zarafa servers isn't. I will check the next time this happens.

I have used Zarafa successfully in several large organizations (with more capable local servers) and have been happy with it. Particularly because it supports activesync for Outlook 2013 and 2016. Not my email client of choice but my wife depends on it for tax and quickbooks work.

I do have the apache web server installed but, I believe, zarafa uses a sandboxed copy, not the general use marketplace install. I can remove that server and will try to see what other marketplace products I can uninstall. The zarafa server was getting heavy use as I moved thousands of emails in multiple folders from local storage to zarafa. That activity might have caused the failure.

I would still like to find out how to stop the zarafa servers mentioned in the original post from starting on reboot. Another option might be to find a way to test if specific servers are running and either start or stop them using a cron job. Will take some research.

Don't use zarafa as it's a resource hog; or more likely the sql database(s) it uses...
However, would consider any program that fails just because swap is needed to be questionable..

Is it the use of swap - or the fact you either don't have a swap space or it's not large enough?

# free -m
or to dynamically monitor
# watch -n 10 free -m

Are you getting any 'invoked oom-killer', 'Kill process' or similar messages in your logs?

An example - my 2G machine ClearOS 7.2 :-

[root@alex ~]# free -m

              total        used        free      shared  buff/cache   available

Mem:           1936        1322         183          21         431         396

Swap:          4095        2272        1823

http://danda.sraellis.tk/frame-28-swap.html