[SOLVED] Cluster disintegration after upgrade to Corosync3

[Ivan Gersi]

I have 5 nodes cluster and I`m going to upgrade to v6. I`ve updated all nodes to latest v5 (5.4-13) adn upgrade Corosync to v3 (because v6 need v3 Corosync).
After upgrade I tried to upgrade node1...it was succesfull to v6 but after rebooting there is some problem with new linux core...but this is not my majority problem.
The huge problem is cluster`s stability because I have 3 repeating scenarios.
1 scenario...cluster working properly, all nodes are visible.
2nd scenario...all nodes are off, no one can see aother one.
3rd scenario...nodes are in semi-visoble mode...I can see another nodes not liek offline (red cross) but like grey questiom mark....after corosync service restart often started turn green again.
I`m little confused/

There is no problem in the network, because these problems stared after upgrade and I have never it before.
Edit: I think this is the same issue like https://forum.proxmox.com/threads/pve-5-4-11-corosync-3-x-major-issues.56124/

 

[Dominic]

Do you still have this problem? Did you upgrade on node to PVE 6 and leave the two others on PVE 5?

 

[Ivan Gersi]

No I resolved this problem with another switch disconnecting from LAN and Corosync is set in separated subnet via this switch. Cluster is stable over one week.
I suppose the main problem of new Corosync are microgaps (short-time latency up) in the switch/network because I have 10Gb network with no problem but Corosync randomly crashed every day. I tried all recommendations....2 rings in every node, separated subnets (without vlan), diferent corosync setup but no progress. I have latest version pof Corosync and libknet library.