pacemaker + podman + compose looks like it might just work

Reply to this note

Please Login to reply.

Discussion

Might actually be something nostr:npub1ecdlntvjzexlyfale2egzvvncc8tgqsaxkl5hw7xlgjv2cxs705s9qs735 that piques your interest. Basically trying to achieve a high availability application environment with minimal moving parts and a smoother learning curve.

Is something like this relevant?

https://github.com/flemay/3musketeers

Doesn't seem like it no, at least not what I'm trying to achieve here. My goal is cloud availability for services just locally.

I mostly use just this nowadays:

https://docs.podman.io/en/latest/markdown/podman-quadlet.1.html

Because at home I typically don't have multiple nodes.

That's just FYI, that very simple setup can be achieved with just systemd.

At work I used pacemaker and It's great.

Podman + systemd and manual deployments have been _fine_ for my business, but I want to be able to run better stuff and offer better guarantees. Mostly so I don't have to be available 24/7 worrying if a service has downtime. Kubes is just too much for me right now and my infra is apparently too complicated to get what I want out of it.

have you got hands on quadlets? I like how simple it is (It's just systemd templating, so elegant)

I have not. The docs (or lacktherof) keep scaring me off. I've really liked compose for the easy of control and yaml I guess. Pacemaker has a built-in integration for compose that is working okay so far. It's an easy transition.

i like that if you understand systemd you're more or less done.

In that case it's probably because I don't fully understand systemd!

nobody fully understand this operating system :)

Probably still worth learning, but not sure how well I can integrate it into a multi-node cluster. It does appear that pacemaker seems to replicate unit files but I'm not fully sure.

It's no use for multinode. at least you'd have to wtite your own "pacemaker"

It's really fun moving containers and floating ip around nodes and having very little downtime.