Trouble with performance tuning (high reported user count and RAM usage)

Technical assistance
1

Infos:

  • Used Zammad version: 6.5.0-1745586583.0f998168.noble

  • Used Zammad installation type: package

  • Operating system: Ubuntu Server 24.04.2 LTS

  • Browser + version: Chrome Version 134.0.6998.198

  • 12 CPU cores

  • 32GB RAM

  • ZAMMAD_SESSION_JOBS_CONCURRENT=4

  • WEB_CONCURRENCY=4

  • ZAMMAD_PROCESS_DELAYED_JOBS_WORKERS=2

  • MIN_THREADS=4

  • MAX_THREADS=16

  • -Xms8g

  • -Xmx8g

  • database - max_connections = 2000

  • 45 agents added, not concurrently online. From the look of the Sessions page in the browser there are only about 20 sessions that have updated in the last hour.

Expected behavior:

  • Smooth performance in browser
  • RAM usage on server not constantly increasing
  • See a realistic output from rails r “p Sessions.list.uniq.count” command
  • Emails processing in a timely manner

Actual behavior:

  • We’re often seeing Error 500 messages pop up in the browser. These can usually just be dismissed without noticeable side effects, but it is concerning and I can’t see it referenced in /var/log/zammad/production.log
  • After already setting ElasticSearch options for -Xms8g and -Xmx8g, after a fresh boot of the server we see about 16GB RAM usage in htop, which slowly, but continuously grows until the server is maxed out. Highest memory user appears to be /usr/bin/redis-server at about 16GB.
  • zammad run rails r “p Sessions.list.uniq.count” provided an output of “1900” even after a fresh reboot with maintenance mode enabled. Documentation indicates that should give a count of current users, but there’s no way we have that many right after reboot.
  • Monitoring page shows 17 emails not processed.

Steps to reproduce the behavior:

  • This is happening consistently across service restarts and server reboots.

I’ve read through numerous threads here talking about performance issues and tried everything I could find. Not sure where else we could be going wrong. Any assistance is appreciated. Thank you!

2

Update: This morning we discovered the drive for Zammad was full, so no one could log in. Only option appeared to be moving to a new host with a larger drive. We setup a new VM, restored a backup, set all the same Env variables and were now back online. Same number of CPU cores, same RAM, but so far we’re now sitting fairly steady around 12GB used. Also, rails r “p Sessions.list.uniq.count” is outputting a much more reasonable number. Not sure what happened, but things seem to be running smoothly now. Will be keeping a close eye on it.