Adding node 5.2 to cluster on 5.0

[Manny Vazquez]

I have a node with 3 nodes (5.0) one of them died, the other 2 took over perfectly fine, now I am ready to bring back to 3rd node, but I did it on 5.2.

Is it ok to add to same node? or should I start a new node and add the other ones as I upgrade them?

 

[wolfgang]

Hi,

it is always better to have the same version in the cluster.
We only test cluster with nodes on the same version level.

I would recommend you to update first your tow nodes and then join the third node to the cluster.

 

[Manny Vazquez]

Hi,

it is always better to have the same version in the cluster.
We only test cluster with nodes on the same version level.

I would recommend you to update first your tow nodes and then join the third node to the cluster.

Thanks, but.. how do you manage upgrading on a cluster? most people can not take down the whole cluster for update, so, shouldn;t the update be done in one node at the time?

This server (node) was part of the cluster, but it died (2 hard drives failed on the raid, what are the chances of that?) so I put new drives and reinstall proxmox, but when I finished, I realized I did it with 5.2.

So, I understand you test with all nodes on same version, and that you recommend I update before, etc.. but how do you solve the paradigm of DOWNTIME to upgrade all the nodes before bringing it up? Each node will take easy one day to upgrade, since I am changing from raid 5 to raid 6 to have better failure support.

 

[wolfgang]

most people can not take down the whole cluster for update, so, shouldn;t the update be done in one node at the time?

Yes update one and then migrate to the other one.

Each node will take easy one day to upgrade, since I am changing from raid 5 to raid 6 to have better failure support.

I do not understand what a version update has to do with Raid.
If you will change the disk layout you can do this later.

 

[Manny Vazquez]

Yes update one and then migrate to the other one.


I do not understand what a version update has to do with Raid.
If you will change the disk layout you can do this later.

Wolfgang, maybe it is lost in translation, english not been the mother language for either one of us.
You recommend that I "update" the other nodes and then ADD this "new" one. But I am not doing an UPDATE to the old nodes but a full reformat, (change RAID geometry at hardware level) and reinstall, this process takes quite a few hours. And I can not have the whole cluster down for 2 days while I do this. That is why I was asking if I could put the new 5.2 in the cluster, move some VMs to it, and take down one other node, reformat, reinstall and put in the cluster, then repeat to the third one.

I understand that I can migrate VMs to the new node, with out this been in the cluster, that implies of course bring down the VMs and (I prefer) restore from backup on the new node. I was hoping that if I could add the 5.2 to the cluster, I could live migrate after doing a replication which is working fine among the other 2 nodes.

 

[Manny Vazquez]

Tried adding the node but I get this error,

what is use_ssh ?

 

[oguz]

what is use_ssh ?

The '--use_ssh' flag on 'pvecm add'. Read 'man pvecm' for more info on it.

You are getting this error because there occurs an exception in the API client (most likely because of the differing versions between node and master). Adding this flag will fallback to the old ssh-only join method (forces the join on the master node)