Hello all, we have recently upgraded to RT5 and now get regular and sporadic out-of-memory errors and have to restart the web service to resolve. Has anyone else experienced this issue or been able to address?
Lots of people are running RT5 in production (including me) so I don’t think this is a universal issue. Could be specific to a feature you use or a customization or plugin? I think I see about 250-300MB of memory usage per RT fcgi process. How much memory is getting used before you have to restart?
Monitor the memory usage of your web service to identify any patterns or spikes. This will help you determine if the memory consumption is gradually increasing or if there are sudden spikes that lead to out-of-memory errors.
If you have (for example) upgraded from CentOS/RHEL 7 to 8 and are using Apache HTTP then your multi-processing model (MPM) is likely set wrong because the OS default changed. RT is only designed to work with the ‘prefork’ MPM but most modern Linux distributions default to the ‘event’ MPM which can cause RT to have severe performance problems including excessive memory usage (see the
web_deployment.pod doc file that came with your RT sources).
i am also looking for the same solution
Thanks for everyone’s input and feedback here! A quick update:
The sysadmin noted some file opening/compilation errors related to RT in the Apache httpd logs. He tracked this to a specific script file line number based on the log:
> return (0, 'no system user', "Couldn't fine RT_System user in the DB '". $RT::Handle->DSN ." ' ");
I’ve never seen an “RT_System” in the RT application user list but when I searched for it directly, I was routed to the “Modify the user RT_System” page. I noted that the “Let this user be granted rights (Privileged)” box was unchecked, which seemed inaccurate for a system-level account. I checked this box and we’ve not experienced any application outages since. Not sure these are 100% related but wanted to share our findings.
Having not seen this user profile before, I am not sure what these settings were prior to the server migration and upgrade to RT5.