pvestatd keeps causing problems with node status on web front-end.

[Eric Renfro]

Hi,

I've recently installed a 4-node cluster of Proxmox 2.0 VE with the HA support (beta 2?) and I keep noticing a reoccurring problem with the reporting on the front-end web interface. First it started with just one of the 4 nodes where it would show as offline, but it would still get the node stats constantly (CPU usage, memory, HDD usage, etc..), but the charts stop, the VM status shows black instead of white even though they are online and all.

When I restart pvestats on that node, it comes back up... For a while, but then it stops again after maybe 10 minutes.

This is starting to happen now on a second node in the cluster, and I fear it's a continuing problem. Checking my logs, I don't see anything related to pvestatd having any issues at all.

What I'm doing now to try to find where the root problem is, I've restarted rrdcached then pvestatd afterwards, but that seems to have totally broken the graphing now, they show up as broken links, and still the nodes in question still went "offline".

Psi-Jack

 

[dietmar]

Please make sure you have the latest packages.

What is the output of:

# cat /etc/pve/.members

In case a node goes offline, does it help if you restart 'pve-cluster' on that node?

# /etc/init.d/pve-cluster restart