Question about stupid cluster tricks....

[mbunds]

I am working within a testing environment, for learning and experimenting before placing Proxmox into live duty. This environment obviously allows experimenting with features and configuration tests that frequently break things, but allow restoration of any part from snapshots, usually within minutes, so I really never have to recover any part of the system by fixing these broken features, except for times when a broken feature presents a troubleshooting path that will be useful for actual maintenance.

So this is merely a curiosity question.

I accidentally placed a duplicate node having a different I.P. than its "twin" (a freshly cloned onto a new server) online with a cluster.

I ignored the odd behavior in the new instances' cluster status (after I convinced it to accept my login) as I realized my mistake and shut-down its original node.

As mentioned above, the original states of the cluster and its nodes can be restored without problems, but I'm wondering if a stupid mistake like this did any damage to the cluster?

I'll be resetting the system in the name of progress and expedience anyway, just in case I did damage it, but I do see the potential to commit the same sin against a production cluster, for exactly the same reason (cloning to a new server), if someone is careless enough to do this.

 

[LnxBil]

I'll be resetting the system in the name of progress and expedience anyway, just in case I did damage it, but I do see the potential to commit the same sin against a production cluster, for exactly the same reason (cloning to a new server), if someone is careless enough to do this.

Why would you do this? Have you done this with your current or previous virtualization environment?

 

[mbunds]

Why would you do this? Have you done this with your current or previous virtualization environment

As I said, this is a test environment being used to model the final production system, which is backed-up redundantly during development so we don't have to care what happens.

Anything done to the cluster, its nodes, and any of their CTs and VMs can be rolled-back within minutes, including grandfathered snapshots of the entire project taken at key "evolutionary" steps.

We can afford to be lackadaisical, cavalier even, because we don't have to care.

There is a solid goal though, and I would never intentionally place an identical node onto the cluster network for any reason I can think of, but it was early Sunday morning, and spinning-up a clone to move to an independent server is a no-brainer, unless lack of caffeine causes you to put it online before you took down its source.
When I realized what I had done, I shut down the source and went to work on the clone because it will be stripped of its cluster configuration anyway, and become a branch for development on an isolated server.

If I corrupted the original cluster or not, I don't care because it has been restored, and now running happily with shares to storage on its former clone.

I just wondered if placing a twin node online, which nobody would do purposely, would certainly corrupt the cluster at large, or would the cluster "heal" from attempting to manage identical resources if the offending node is removed.

Simply curiosity.

 

[LnxBil]

early Sunday morning [...] lack of caffeine

Okay, got it

I just wondered if placing a twin node online, which nobody would do purposely, would certainly corrupt the cluster at large, or would the cluster "heal" from attempting to manage identical resources if the offending node is removed.

Nice hypothetical and I really don't know - with any cluster technology what would happen - I assume chaos.

 

[mbunds]

Okay, got it


Nice hypothetical and I really don't know - with any cluster technology what would happen - I assume chaos.

Thanks!

I would never tempt fate purposely, and we could do the same thing purposely and study the results, but there are deadlines (which is why the project is so redundantly backed-up), so no time to play; I was just curious.

I got the same question from partners; "You put a cloned node online with is source? Why the hell would you want to do that!??

I can say, that the cluster manager was reacting very strangely, but I didn't leave this running long before I realized what happened, then rolled-everything back.

 

[LnxBil]

I got the same question from partners;

Ask they also what happens if you plug in your mains into your <insert hardware part here> (e.g. a ethernet killer)? There are much easier ways to f**k up your network than that.