Seeking advise - 2 nodes, to cluster, or run independently?

[Tigo]

I have two nodes currently running independently (production - stable / lab - testing), - each with its own local storage for backup. I am thinking of placing them in a cluster, in order to:

1 - Be able to run a backup job from the production node to the lab node storage.
2 - Be able to replicate vms - lxc - as a back up from production to lab node.
3 - Be able to create new vms - lxcs on the lab node, and then once ready to push them out to the production node.

I think that I should be able to do the above 3 points. But, I'm concerned about:

1 - should I shutdown the lab-node, - will that cause any issues for the production node? And vice-versa, Production shutdown, any issues with lab node?
2 - Will I able to recover the Production node (inclusive of vms & lxcs) should it crash (dead ssd), by restoring the backups from lab?

Thank you,

 

[bbgeek17]

I have two nodes

Two-node cluster is not something you should pace in production

1 - Be able to run a backup job from the production node to the lab node storage.

you dont need a cluster for this

2 - Be able to replicate vms - lxc - as a back up from production to lab node.

if you already have a backup, you dont need to replicate again. You can use underlying replication directly (assuming you use ZFS). If you dont use ZFS, then you cant replicate even within a cluster.

3 - Be able to create new vms - lxcs on the lab node, and then once ready to push them out to the production node.

You can use new cross-cluster/node copy functionality for this

1 - should I shutdown the lab-node, - will that cause any issues for the production node? And vice-versa, Production shutdown, any issues with lab node?

With two node cluster you will definitely have issues. About 30% of forum posts are related to this exact situation.

2 - Will I able to recover the Production node (inclusive of vms & lxcs) should it crash (dead ssd), by restoring the backups from lab?

In general yes. However, it depends on what you mean by backup. Recovering the underlying hypervisor, at this point, is left to administrator's imagination. I.e. its not integrated into PVE, yet.

Good luck

---

Blockbridge : Ultra low latency all-NVME shared storage for Proxmox - https://www.blockbridge.com/proxmox

 

[Tigo]

you dont need a cluster for this

Can you refer me please to the document on how to do the backup from production to the lab node? (do I just setup a backup server in a VM on the lab node, and then a client on the production node? or?

if you already have a backup, you dont need to replicate again. You can use underlying replication directly (assuming you use ZFS). If you dont use ZFS, then you cant replicate even within a cluster.

Am using ZFS. I'll need to look up this replication aspect - didn't pick up on it when I was setting things up.

You can use new cross-cluster/node copy functionality for this

hmmm, another point that I wasn't aware of - didn't note it. I was backing up containers / vms, and transferring them from the lab node to the production node, then restoring after making some adjustments. Need to revisit the documentation.

With two node cluster you will definitely have issues. About 30% of forum posts are related to this exact situation.

Sounds good - best to avoid then.

In general yes. However, it depends on what you mean by backup. Recovering the underlying hypervisor, at this point, is left to administrator's imagination. I.e. its not integrated into PVE, yet.

As the administrator of my home environment. Need to think of an approach then. Production node simply runs the household server, and then the lab is where I can build at before pushing it out.

Good luck

Many thanks

 

[RobGranger]

If anything happens to one node your backups will fail. I had all kinds of issues with 2 node, I finally broke it and just do backups from shared NFS storage on each Truenas server. I cross backup so the backups are on the other server.

 

[Kingneutron]

@RobGranger did you have any consideration for implementing a Qdevice? 2-node clusters are known problematic