One way only replication.

[Francesco P.]

Hi,

I have a 2 node PVE (with a third corosync quorum node), with replication configured.

I'd like to have one way replication only, so that when node A is active, it does replicate on node B, but when node B became active, it does not replicate VMs back to node A. That's because replication is done only 3 times a day, and sometimes node A restarts for still unknown reasons: in that event I don't want node B to automatically take over, I want to check what's going on before restarting things.

I've disabled all resources in HA, so that VMs are not automatically migrated and restarted in case of node A failure; is that enough to avoid also "reverse" disk replication?

Or is there some other way to achieve this behaviour?

Thanks!

 

[jmckee]

You must be a small business person like me. Maybe you can tell by the silence that they are not interested in you.
That being said, your use case seems a little unusual.
I don't think you need to worry about unwanted failover if you have HA disabled. Without failover then there is no need worry about "reverse" replication. Back in the days when drbd worked there was the concept of primary and secondary. In that context Node A is your primary and Node B is secondary.
These days I am not really sure what is recommended for small clusters... my HA has been broken for a long time. I probably need to learn to use ceph. That leads to the next comment - you don't describe your setup in much detail. What technology are you using?
Good luck

 

[Francesco P.]

Yes, I'm a micro-business, with small customers.
And yes, disabling HA also disable the two-way replication, I checked that in the meanwhile.
All I know about Ceph is that it requires (well, it is strongly suggested) at least three nodes, and a lot of expertise. For small installations like mine, maybe some kind of NAS or external storage would be more reasonable.
What do you mean by "technology"? :? They are two phisycal server on Hetzner, with PVE 7.2, and nothing much more!

 

[jmckee]

I mean how were you accomplishing replication? There are different ways (technology).
Two servers is enough to provide redundancy and failover but you need a third to provide "quorum".

 

[Francesco P.]

Ah yes, I made the third - quorum only - node following this guide:

https://florianmuller.com/setup-a-p...berrypi-as-a-third-quorum-vote-device-qdevice

But I used a small VM instead of a rasperry. Then everything works with the standard PVE tools.