pvescheduler memory leak on PVE 9.1.6

H

hpahnke

New Member
Mar 10, 2026
3
0
1
Hi all,


I've noticed a significant memory leak in pvescheduler on my Proxmox VE 9.1.6 installation. Here are the details:


Environment:


  • proxmox-ve: 9.1.0
  • pve-manager: 9.1.6 (running version: 9.1.6/71482d1833ded40a)
  • proxmox-kernel: 6.17.9-1-pve
  • Single node, no cluster
  • 6 LXC containers being backed up daily via vzdump

Symptoms:


  • After a fresh restart, pvescheduler uses ~120MB of RAM
  • Within 24 hours it climbs to 10-15GB
  • Over 8 days it reached a peak of 24.6GB
  • During backup jobs it spikes even higher (~15.3GB peak observed)
  • Memory pressure stalls visible in the Proxmox dashboard during backup jobs
  • Eventually pvescheduler becomes too degraded to write /etc/pve/vzdump.cron, causing scheduled backups to silently not run
  • Restarting pvescheduler immediately resets memory back to ~120MB

Workaround:I have a cron watchdog restarting pvescheduler when memory exceeds 17GB, but this is not a permanent fix.


Has anyone else experienced this? Is there a known fix or patch in the works?


Thanks
 
Update: A colleague running PVE 9.0.3 reproduced the same issue. After backing up a single small LXC, pvescheduler went from 117MB to 440MB (peak 1.7GB) and did not release memory after completion. This confirms the issue spans multiple versions and is not specific to 9.1.x.
 
What values did you look at and where? I assumed you used sometihing like top -co%MEM.
 
You must log in or register to reply here.
Top Bottom
Back