Updating all Proxmox Nodes Simultaneously

[Lonnie]

I have Proxmox 3.3 cluster, and I've been updating each node individually using the web GUI.

I've noticed that when I update one node, it takes a little time before the GUI will allow me to update another node. The NoVNC console fails to showcase the update-command-line if I try to update a node too soon after I've just gotten through updating a previous node.

I'm wondering why a certain amount of time must pass between updating each node. Also, I'm wonder how people (who have a lot more nodes than me) perform node updates. Are there methods for propagating updates to all nodes at once?

For me, it is not too much trouble to update each node individually, but I'm curious how administrators with several nodes administer updates, and I'm also curious as to why I must wait a little while between each node update (via the web GUI).

I'd appreciate your insight. Thanks.

 

[e100]

I've never used the GUI to update nodes so I have no idea why you need to wait between updating nodes.

As far as how I manage updating lots of nodes, I use Chef and SSH to accomplish the task.
If you are not using Chef you can still just use SSH.

With Chef I simply run a single command like this to update all of my Proxmox nodes:

knife ssh "chef_environment:production AND os:linux AND fqdn:proxmox*.domain.com" "sudo apt-get update && sudo DEBIAN_FRONTEND=noninteractive apt-get dist-upgrade -y -q"

With only SSH I would suggest logging into one Proxmox node and as root run something like this for each node:

ssh OtherProxmoxNode.domain.com "apt-get update && DEBIAN_FRONTEND=noninteractive apt-get dist-upgrade -y -q"

 

[Lonnie]

I just wanted to report back, that I'm no longer having to wait between node updates; lately I've been able to update one node right after another in the web-gui without issue.

Additionally, thank you e100 for sharing how you automate updates of all nodes at once.

 

[mo_]

you might also want to look at clusterssh/cssh: its basically input multiplexing into multiple ssh sessions. i.e. you can do the same mistake/typo on all your systems at once!

 

[e100]

Ha!
I know I've done that more than once.

 

[shantanu]

In case you want to save some time, one can rsync the apt cache from the first node
onto the next one.

The upgrade on the second node will thus not download the data again!

I always do this, works great for me!

Regards,
Shantanu Gadgil

 

[e100]

I've never thought of using rsync to copy the apt-cache.
To speed things up I installed apt-cacher-ng on one node.

Then on all the other nodes added the following to /etc/apt/apt.conf.d/01proxy

Acquire::http::Proxy "http://cache.domain.com:3142";
Acquire::https::Proxy "DIRECT";