Very long HA

[Akhyo]

Hello,

I have a Dell C6220 server that hosts 3 nodes with configuration:
- Bi Xeon E5-2630L
- 32 GB RAM
- 1x500Gb (Proxmox 5.1-35) and 3x1To (Ceph)
- 1G link in private

This is just an installation to test the functioning of the CEPH and HA, so I created a cluster with the 3 nodes, I then used 9 disks (3 per server) to create a CEPH, which works perfectly. I then set up the HA by creating a group linking the 3 servers, and when I create a container on one of the nodes and cut it off, the automatic migration takes almost 5 minutes!

Is that a normal delay? Would it be possible to fix this? Or would it be a configuration error?

With kind regards

 

[wolfgang]

Is that a normal delay? Would it be possible to fix this? Or would it be a configuration error?

This is the normal maximum time and this time is needed to be sure all required conditions are true.
You have to wait longer than the self fencing is to ensure that the node is down.
Than if your unlucky and the node was the master of HA then a new election will start.
This election need an other waiting time to free the masterlog.
After this the migration can be started.

And no you can't make this waiting times shorter.