[solved] Cluster has quorum but trouble with pvestatd/memory and /etc/pve

udo

Distinguished Member
Apr 22, 2009
5,977
199
163
Ahrensburg; Germany
Hi,
due to memory-problems (to much need during backup?!) one node was red in the web-gui, but the nodes has quorum:
Code:
pvecm status
Version: 6.2.0
Config Version: 3
Cluster Name: cluster
Cluster Id: 62900
Cluster Member: Yes
Cluster Generation: 296
Membership state: Cluster-Member
Nodes: 3
Expected votes: 3
Total votes: 3
Node votes: 1
Quorum: 2  
Active subsystems: 5
Flags: 
Ports Bound: 0  
Node name: proxmox1
Node ID: 3
Multicast addresses: 239.192.245.170 
Node addresses: 172.20.x.61
pvestatd don't run and a restart has now the issue - if I look on this node I see only this node green in the gui, and from both other nodes I see the other nodes green.

I found (hope so) the beginning of the trouble:
Code:
Oct 12 21:49:07 proxmox1 pvestatd[7283]: WARNING: command 'df -P -B 1 /backup' failed: open3: fork failed: Cannot allocate memory at /usr/share/perl5/PVE/Tools.pm line 280
Oct 12 21:49:07 proxmox1 pvestatd[7283]: WARNING: command 'df -P -B 1 /mnt/local_iso' failed: open3: fork failed: Cannot allocate memory at /usr/share/perl5/PVE/Tools.pm line 280
Oct 12 21:49:07 proxmox1 pvestatd[7283]: WARNING: command '/usr/bin/iscsiadm --mode session' failed: open3: fork failed: Cannot allocate memory at /usr/share/perl5/PVE/Tools.pm line 280
Oct 12 21:49:07 proxmox1 pvestatd[7283]: WARNING: command '/usr/bin/iscsiadm --mode session' failed: open3: fork failed: Cannot allocate memory at /usr/share/perl5/PVE/Tools.pm line 280
Oct 12 21:49:07 proxmox1 pvestatd[7283]: WARNING: Use of uninitialized value $storeid in hash element at /usr/share/perl5/PVE/Storage.pm line 680.
Oct 12 21:49:07 proxmox1 pvestatd[7283]: WARNING: Use of uninitialized value $storeid in hash element at /usr/share/perl5/PVE/Storage.pm line 681.
To get more ram, I would migrate an VM but get the error:
Code:
root@proxmox1:/# qm migrate 114 proxmox3 
Oct 15 13:48:11 starting migration of VM 114 to node 'proxmox3' (172.20.x.63)
Oct 15 13:48:11 ERROR: unable to open file '/etc/pve/nodes/proxmox1/qemu-server/114.conf.tmp.51479' - Device or resource busy
Oct 15 13:48:11 aborting phase 1 - cleanup resources
Oct 15 13:48:11 ERROR: unable to open file '/etc/pve/nodes/proxmox1/qemu-server/114.conf.tmp.51479' - Device or resource busy
Oct 15 13:48:11 ERROR: migration aborted (duration 00:00:01): unable to open file '/etc/pve/nodes/proxmox1/qemu-server/114.conf.tmp.51479' - Device or resource busy
migration aborted
/etc/pve isn't readonly but I can't touch anything:
Code:
root@proxmox1:~# touch /etc/pve/xx
touch: cannot touch `/etc/pve/xx': Device or resource busy
root@proxmox1:~# mount | grep /etc/pve
/dev/fuse on /etc/pve type fuse (rw,nosuid,nodev,default_permissions,allow_other)
Any hint?

Udo
 
Last edited:
Re: Cluster has quorum but trouble with pvestatd/memory and /etc/pve

You run out of memory, so arbitrary services are crashing. You first need to stop something too free memory - then cleanly restart failed services.

Any idea why you run out of memory? (I just fixed a memory leak in pvestatd today).
 
Re: Cluster has quorum but trouble with pvestatd/memory and /etc/pve

You run out of memory, so arbitrary services are crashing. You first need to stop something too free memory - then cleanly restart failed services.

Any idea why you run out of memory? (I just fixed a memory leak in pvestatd today).
Hi Dietmar,
can be pvestatd - i have only one VM stopped (1GB) and now I have 4GB free...
But in spite of restarting pvedaemon and pvestatd the node is red but pvecm say all ok?!

pmxcfs give following error:
Code:
Oct 15 18:24:54 proxmox1 pmxcfs[6722]: [status] crit: cpg_send_message failed: 9
Ahh - just do an
Code:
/etc/init.d/pve-cluster restart
and all nodes are green!!

Problem solved - so til pve2.2 it's recommended sometimes restart pvestatd?

Thanks - Udo
 
Re: Cluster has quorum but trouble with pvestatd/memory and /etc/pve

But in spite of restarting pvedaemon and pvestatd the node is red but pvecm say all ok?!

pvecm displays cluster status (cman/corosync), and pvestatd send node/VM status to
other nodes (which is diplayed by GUI).