Adding new Node to Cluster

[andy77]

Hello @ all,

I just want to be sure about the procedure to add a new node to a cluster. The cluster is actually running 4.2-18. So the first question would be, can I add a 4.4 node to a 4.2 cluster? Or do I need to install the node with 4.2, or maybe first upgrade the cluster to 4.4 and then add the new node with 4.4 also?`

Thx for any help

 

[Ashley]

Either upgrade and then add node or install node with the exact same version.

However as there is no reason to not be running the latest PVE version I would say to upgrade first.

 

[andy77]

Hello, ok thx that what I though. But when I have a cluster of 4 nodes, where to start upgrading? (at the "Master node" ?)

 

[Ashley]

No specific order, just make sure you do apt-get dist-upgrade instead of just apt-get upgrade.

This will have no impact to any running VM's, however your find probably will bring along a new kernel upgrade so a boot into this when you can do would be a good idea.

 

[andy77]

Hmm, but in the Proxmox Wiki is is written

Before you upgrade your system, you should stop all your running VM´s and LXC containers.

 

[Ashley]

What page within the Wiki did you see that? Have done multiple upgrades with no issues.

 

[andy77]

https://pve.proxmox.com/wiki/Downlo...Proxmox_Virtual_Environment_4.x_to_latest_4.4

 

[fabian]

Hmm, but in the Proxmox Wiki is is written

http://pve.proxmox.com/wiki/Downloads#Update_a_running_Proxmox_Virtual_Environment_4.x_to_latest_4.4 <- I don't see this? Maybe you are referring to the instructions for upgrading 3.4 to 4.x ?

 

[Ashley]

As fabian I do not see this on the link you sent, and is defiantly not required.

 

[andy77]

No, ist is for sure in the 4.x to 4.4 update. Here the whole row:

If you get any errors, your sources.list (or your network) has a problem. Before you upgrade your system, you should stop all your running VM´s and LXC containers.

 

[Ashley]

That's a mistake in the WIKI most probably a copy and paste mistake, can safely be ignored.

Edit - as below an upgrade on its own won't touch any running VM's however without a reboot / migration out and back in you won't benefit from new kernel or QEMU versions.

 

[fabian]

No, ist is for sure in the 4.x to 4.4 update. Here the whole row:

missed that I think that is more of a recommendation - if it is possible to migrate / stop them, it's better. but it is not needed in 99% of the cases. in many cases you'll need to reboot anyway because of the kernel upgrade..

 

[andy77]

Ok, no other recommendations for upgrade? Sounds to me "too easy" ;-)

 

[andy77]

BTW: When I start to upgrade one of the nodes, and then Irestart it. Will it still work in the cluster (that is 4.2)? Or do I have to upgrade all nodes at the same time.

And again, to be 100% sure, I do not have to be aware of the "cluster master node"? To update it first, or last, or whatever?

 

[Ashley]

It will work fine as a cluster, however will be unable to do things like live migration between nodes until you have run apt-get dist-upgrade on all due to PVE version miss match.

I would say upgrade one reboot to make sure no issues with new kernel / grub on your hardware. Then upgrade all other nodes to latest and do live migration / reboots as required (if supported storage used)

 

[fabian]

BTW: When I start to upgrade one of the nodes, and then Irestart it. Will it still work in the cluster (that is 4.2)? Or do I have to upgrade all nodes at the same time.

the clustering itself should not make any problems when upgrading individual nodes (the core technology used is very stable). sometimes there are incompatibilities regarding the live migration of VMs, those can usually be resolved by upgrading the packages (without rebooting), then doing the live-migrate, then doing the reboot. migration from new to old is not supported in general, so you should upgrade the nodes one after another with appropriate shuffling around of guests to restore your cluster's full capabilities.

And again, to be 100% sure, I do not have to be aware of the "cluster master node"? To update it first, or last, or whatever?

there is no master node PVE is fully distributed.

if you are unsure about any of this, you can recreate your cluster setup either on test hardware or in a virtual environment (with CPU host you can have nested virtualization) and just attempt the upgrade in such a safe environment first. I'd recommend having a test environment resembling the real one as closely as possible for any production setup anyhow..