Infos:
- Used Zammad version: 5.2.0
- Used Zammad installation type: debian package
- Operating system: debian 10
- Browser + version: different
- ESXi, VM has 8 cores E5-2697a V4, 40GB RAM, SSD backed storage, VM host is not too busy…
- HAProxy as reverse proxy in front of nginx (certificate handling etc)
Expected behavior:
- zammad works nicely and smoothly
Actual behavior:
- zammad is slow, opening tickets takes 20-30-40 seconds
- high rails server and websocket-server load (see htop screenshot)
Steps to reproduce the behavior:
- have 10-20 agents online
- log in
- wait
After updating from build 80bf57c1 (we had 1 or 2 ruby/puma processes running at 90-100% all the time and one background-worker.rb with 50-90%) to 743b9821 (I luckily found After updating to 5.2.0 version Zammad works very slow - Technical assistance - Zammad - Community half an hour ago) it got better, puma does not produce as high load. Still not too desirable though.
performance optimizations:
- elasticsearch.yml:
http.max_content_length: 400mb
indices.query.bool.max_clause_count: 2048
We are running 10-30 agents simultaneously, and unfortunately there are ~40 overviews (which are restricted group-wise, so each user has a maximum of 10 overviews, less is not possible unfortunately)
Any clue how to get this to work smoother again? Before we updated to 5.2, it was really quick…