HA and manual vms starts

[dmvillar]

hi !

I have a cluster of 3 nodes, where node1 is where I run several vms, and I replicate data every 1 hour to node2 (do not configure shared storage because the network is only 1gb)

What I am looking for is that when node1 fails, it allows us to review the failure before starting the vms on node2 (we would start the vms on node2 only in case of a severe failure on node1)

I have tried different HA configurations, but I cannot achieve this behavior.

Is there any configuration available for this?

 

[dmvillar]

I think I found it!

I'm sharing it, in case anyone is looking for something similar.

1. Configure replication every X minutes/hours from node1 to node2.

(To have a copy of the recent disk, in the event of a serious failure on node1)

2. Configure HA on each VM with "Request State" = "Ignored"

(Given the "non-response" of node1, first check what is happening with node1, and if possible, quickly solve and start node1)

3. If node1 is severely damaged and it will take too long to get it working again, enter the cluster from node2, go to > Datacenter > HA and CHANGE "Request State" = "started".

(The vms and cts will automatically start on node2, this depending of course on the group configuration)

If anyone sees any improvements to this process, I would really love to read it.

thank you really

 

[UdoB]

What I am looking for is that when node1 fails, it allows us to review the failure before starting the vms on node2 (we would start the vms on node2 only in case of a severe failure on node1)

Keep it simple, stupid: just do not enable HA for that VM... ;-)

 

[dmvillar]

Keep it simple, stupid: just do not enable HA for that VM... ;-)

Yeah ! It is something that I have tried, but in case I need to move the 80vm to node2, because node1 is literally dead, there is no way to move it without doing hard work from the console, this way just by changing "Request State" in HA the vms They travel to node2 without doing anything else

 

[Falk R.]

Yeah ! It is something that I have tried, but in case I need to move the 80vm to node2, because node1 is literally dead, there is no way to move it without doing hard work from the console, this way just by changing "Request State" in HA the vms They travel to node2 without doing anything else

The hard work is a simple move .conf files from one folder to another folder.

 

[dmvillar]

The hard work is a simple move .conf files from one folder to another folder.

Yes, one by one 80 config files (at a critical moment), isn't it easier and less prone to error, to change the state from the web UI?

In any case, it is a matter of individual preference.

 

[dmvillar]

In my opinion, the less the console should be used, the more easily anyone can do it just by leaving it documented, even for non-Linux users.