[Operators] Ejabberd upgrades and mnesia multinode setup
Mr. Glen J. Loftin
glenloftin at loftintechnologies.com
Wed Aug 5 23:25:02 UTC 2015
What do you all know about x-ass
Sent from my iPhone
On Aug 3, 2015, at 10:26 AM, Micka?l R?mond <mremond at process-one.net<mailto:mremond at process-one.net>> wrote:
Holger Baust said:
I am planning to upgrade jabber.freenet.de<http://jabber.freenet.de> soon.
But this is a multinode setup with 2 nodes.
Has anyone done this yet and can give me some advises?
(only ejabberd specific please)
Since these Nodes run as Virtual Machines on VMWare clusters,
I have the advantage of using snapshot for roll-backs.
It should be quite a straightforward operation.
My advice are the following:
1. Backup Mnesia DB and relational backend.
2. Take a snapshot of your VM for rollbacks.
3. Assess if you can migrate one node at a time. If you can simulate cluster update on snapshot outside of the production context it would be good. Depending on which version you are and which version you are getting two, it may be possible to mix cluster nodes from two versions (if internal data structure between the two versions did not change)
4. Shutdown nodes (or only one if possible) and upgrade code, but keep data and restart the nodes in reverse order (otherwise, the service will wait for the latest online master to go up to really start).
Does it make sense ? Do you have more specific questions ?
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Operators