[solved] Zammad memory leak?

Hi Nino,

once again: Sorry for my late response. And sorry for the confusion. To be honest I wasn’t directly involved in the task and these are all the information I could gather. To provide you a proper answer I’d have to invest quite an amount of time I currently don’t have due to the Knowledge Base development and support for our customers. However, I’ll provide information as soon as I can (which might take quite a while). Not usual as with proprietary software, we finance Zammad through service. So it’s less money we have at our disposal to push Zammad.

If you would like to support Zammad, you can do so via Zammad Support: https://zammad.com/pricing#selfhosted

Thank you for your understanding.

Hi Thorsten,

Thank you for your elaborate response.
It seems it was a misunderstanding on my part then that the setup can “easily” be fixed.
I can totally understand that you have to tend to your paid projects first.
In the meantime we will see if we can find a solution for our selves with some trail&error and keep you posted as well.
Should you find the time to investigate this further yourself, any hints are welcome.

Best regards,
Nino

2 Likes

Little update.
We reduced now WEB_CONCURRENCY to 3 and MAX_THREADS to 16.
This seems to have drastically reduced memory usage while keeping the speed benefits from earlier.
After 5 days testing instead of an overflow we are currently hovering at around 58% active memory usage.

We will keep tracking this a bit longer and see if we run into memory issues again or if we are good now.
Keep you posted.

2 Likes

OK we seem to be stable now. After 13 days uptime we hover at around 65% RAM usage of the 16 gig.
You can mark this thread as solved now.
Finally we can give the memory hungry elastic search a bit more RAM to play with :wink:

BR,
Nino

Hey there,

thank you very much for your Feedback, glad you could bring it to stable!

Hi Nino! Thanks for your feedback. I’m glad that you managed to find a way - kudos for that! I just read a great article about the memory consumption of Ruby applications. They already created a testable patch that probably will get evaluated over at Github, Discourse etc. We’ll keep an eye on that.
However, the article mentions a workaround by setting the ENV MALLOC_ARENA_MAX=2 (for the session of the Ruby process). This should reduce the memory consumption a lot but will come with the cost of higher CPU usage. We haven’t tested it yet because the memory consumption of our hosted instances is quite good. Anyhow, I just wanted to share the knowledge in this context so maybe someone can profit from it.

This topic was automatically closed 120 days after the last reply. New replies are no longer allowed.