Multicast

[ejc317]

So when I used to do SSMPING, the server returned multicast. Now I just tested again - all of a sudden it shows unicast. However, on the switch side NOTHING has changed and in fact if you do netstat -g, it shows that it is part of a multicast group - any ideas? The switches are Brocade Fast Iron FESX-448-PREM - IP Multicast is active globally

Now I don't lose quorum (pvecm nodes) shows the right nodes. The only issue is that the node details are gone and i have to manually restart pvestatd and apache2 all the time.

One thing to note is that there are 2 proxmox clusters on this network and when 1 cluster is down the other one works without issue. This is quite strange and only started yesterday. should I have separate multicast groups for the clusters?

Interface RefCnt Group
--------------- ------ ---------------------
lo 1 all-systems.mcast.net
eth0 1 all-systems.mcast.net
eth1 1 all-systems.mcast.net
eth2 1 all-systems.mcast.net
eth3 1 all-systems.mcast.net
vmbr0 1 all-systems.mcast.net
vmbr1 1 239.192.36.60
vmbr1 1 all-systems.mcast.net

 

[dietmar]

One thing to note is that there are 2 proxmox clusters on this network and when 1 cluster is down the other one works without issue. This is quite strange and only started yesterday. should I have separate multicast groups for the clusters?

Multicast address should be different if you use different cluster names. Check with

# pvecm status

 

[ejc317]

Multicast address should be different if you use different cluster names. Check with

# pvecm status

Sorry misspoke - multicast address is different

 

[ejc317]

One semi unrelated issue (or maybe related) - its been failing to log into 1 of our NFS share (the openfiler one) and when I removed the NFS share, the pvestatd stays up for longer ...

 

[dietmar]

One semi unrelated issue (or maybe related) - its been failing to log into 1 of our NFS share (the openfiler one) and when I removed the NFS share, the pvestatd stays up for longer ...

Your storage needs to be online - else pvestatd can run into timeouts.

 

[ejc317]

Your storage needs to be online - else pvestatd can run into timeouts.

I've removed the dead storage and added a 15 minute cronjob to execute across all nodes "service pvestatd restart" - this works for now ... but heh not ideal - its also run into issues with multipath

 

[dietmar]

and added a 15 minute cronjob to execute across all nodes "service pvestatd restart"

That should not be needed - do you use latest packages?