Feature Request: Emergency Start

I’ve had ongoing niggly problems with LXD Clustering, and one thing that would save me a bunch of time and heartache is an ‘emergency start’ feature.

Let’s say that for whatever reason your cluster won’t start - there’s plenty of reasons why it won’t, a stalled snap upgrade, a missing node, a database corruption, those sorta things.

What would be amazing is a cluster-destructive ‘I don’t care what you think, start all the containers that are on this machine’ feature. My problems yesterday could have been resolved by this, as the cluster database was waiting for a machine that didn’t exist to start.

If we could somehow go ‘You are the only cluster member left. You are authoritative. Start everything, become a manager, whatever you need to do to ACTUALLY BRING THE CONTAINERS UP’, that will make DR a pile easier.

Is there any chance of that being a thing? That would at least let me start to trust the clustering a bit more, because at the moment I don’t find it trustworthy at all 8-(

1 Like

Yes, we do have planned work for @freeekanayaka to add a self-service database tool which would allow for forceful removal of cluster nodes, reduction of quorum, … so that you can effectively run the machine standalone. You wouldn’t be able to cluster it back afterwards so you’d likely need to setup a clean cluster and move your containers, but that may still be valuable for a number of users.

1 Like