Storage replication require all nodes join cluster or not?

[itvietnam]

Hi,

I have a few VMs and need replicate data to remote datacenter, i seen this articles: https://pve.proxmox.com/wiki/Storage_Replication

If we use this method, does it requires all node must join same cluster? i don't see any requirement listed in above link.

For alternative option, i see other methods such as rbd-mirror and pve-zsync. For rbd-mirror, it's handle at storage level, how about pve-zsync, does it require all node join to the same cluster?

If all above solutions required all nodes join cluster, how stability for remote datacenter? what happen if cluster network (corosync) die for few minutes and may up to few hours to recover.

 

[Alwin]

For alternative option, i see other methods such as rbd-mirror and pve-zsync. For rbd-mirror, it's handle at storage level, how about pve-zsync, does it require all node join to the same cluster?

Use the pve-zsync, as it does a zfs send/receive independently of the cluster. This only syncs the data to remote location, the configs need to be done by an alternative means.

 

[itvietnam]

Thanks for a hint @Alwin, i will check on it.

 

[itvietnam]

I see the section recovery a VM: https://pve.proxmox.com/wiki/PVE-zsync#Recovering_an_VM

How we can ssh to source node to stop the sync job and start recovery a VM due fo disaster and source datacenter is unable to access (like burning, flood...)?

If we can not access to source DC, is there a way to recover VMs with pve-zsync?

And this step is also need to rewrite VM config file to specific storage pool, so we can do this with rbd-mirror?

 

[Alwin]

That's what I meant with, alternative means for the config. If the remote host by any chance is also a Proxmox VE, then you could create the same storage and copy the config into the /etc/pve/ directory. With a little bit scripting you could even create VM/CT through the API on both nodes automatically.

 

[itvietnam]

I see the section recovery a VM: https://pve.proxmox.com/wiki/PVE-zsync#Recovering_an_VM

How we can ssh to source node to stop the sync job and start recovery a VM due fo disaster and source datacenter is unable to access (like burning, flood...)?

 

[Alwin]

How we can ssh to source node to stop the sync job and start recovery a VM due fo disaster and source datacenter is unable to access (like burning, flood...)?

When the source is not available then the pve-zsync job is neither. The docs talk about a recovery, when the source node comes back alive again. But you also can configure the job on the backup node and let pve-zsync pull.

 

[itvietnam]

Hi @Alwin, we talk about recovery in the case of failure happen. For instances, we have host A and host B. Host A configure to sync all VM disk to host B.

Your document mentioned: "You must recover the VM or dataset manually.... Fist, stop the sync job for the VM or dataset in question."

The sync job was created on host A. In this case, host A die, so how we can ssh to this node and stop the sync job? and doing zfs send from host A to host B?

Below command tell me this must handle from host A. But it die and can not get back online, so how we can handle this recovery without host A?

zfs send <pool>/[<path>/]vm-<VMID>-disk-<number>@<last_snapshot> | [ssh root@<destination>] zfs receive <pool>/<path>/vm-<VMID>-disk-<number>

 

[Alwin]

The sync job was created on host A. In this case, host A die, so how we can ssh to this node and stop the sync job?

Ask yourself, when the host is dead, what is the job then?!

and doing zfs send from host A to host B?

It's the other way around.