I lost quorum in cluster (without touch). I dont know why, because only 3 nodes from 16 are affected with this problem, and after reboot affected nodes, cluster goes to hell. Corosync processes growing in memory to full utilization of available ram (64gb ram on server, and corosync process have...
I have same problem, but my cluster (with 16 nodes) has broken and i must restart nodes with 2.6.32-37 kernel. After reboot broken nodes i must restart cman and corosync on all nodes. That is problematic when cluster have 300+ virtuals.
This problem becomes very often, when i running on kernel...
Create dir in for example /mnt/temp-root
mount /dev/pve/root /mnt/temp-root
du -shx /mnt/temp-root/*
The most situation, when is at to root mounted another file system like pve-data has no reporting data "under" the mount.