search

siderolabs / omni

SaaS-simple deployment of Kubernetes - on your own hardware.
Other
479 stars 26 forks source link

`talosctl dashboard` doesn't work when using Omni and node uuid/nodename as `-n` #491

Closed Unix4ever closed 1 month ago

smira commented 1 month ago

I think this got already fixed, at least I remember Utku doing a fix for this case.

Unix4ever commented 1 month ago

I think this issue is different. In talosctl you pass -n talos-vpm-88g, it gets resolved into 10.5.0.2 in Omni and works, then it returns back to talosctl with Hostname=10.5.0.2 and it breaks the dashboard as it tries to find it by the name talos-vpm-88g. I'm not sure if there's an easy way to fix it, but if we want to keep talosctl generic then we need to swap Hostname to talos-vpm-88g in the response.

smira commented 1 month ago

I think that was exactly the fix, it tried to do one API call with each node and captures resolved address back from Talos API.

utkuozdemir commented 1 month ago

I thought I fixed it that way, could it be a regression by the later changes on the name resolution code? In any case, I'll have a look.

Unix4ever commented 1 month ago

I've checked that on the latest Talos, the problem seems to be there.

utkuozdemir commented 1 month ago

This issue does not happen when talosctl and Talos versions are compatible.

However I indeed found an issue on the Monitor tab, which is now addressed by: https://github.com/siderolabs/talos/pull/9099

utkuozdemir commented 1 month ago

I verified one more time, on my demo instance: the problem is not there on Talos 1.8.0-alpha.1. With the issue I fixed, dashboard is fully functional: https://github.com/siderolabs/talos/pull/9099

Hence, closing this.

smira commented 1 month ago

I verified one more time, on my demo instance: the problem is not there on Talos 1.8.0-alpha.1.

Is there any fix in 1.8 which didn't make its way into 1.7?

utkuozdemir commented 1 month ago

I verified one more time, on my demo instance: the problem is not there on Talos 1.8.0-alpha.1.

Is there any fix in 1.8 which didn't make its way into 1.7?

There was an issue with the top section of the overview tab not loading on 1.7.4, but I assumed it was caused by cli<>server incompatibility, but let me dig a bit further to find it out exactly.

utkuozdemir commented 2 weeks ago

I verified one more time, on my demo instance: the problem is not there on Talos 1.8.0-alpha.1.

Is there any fix in 1.8 which didn't make its way into 1.7?

Just checked.

The newer talosctl version (1.8.0-alpha...) was not working properly with older Talos versions (1.7.x) due to some resource types not being available, which I addressed here: https://github.com/siderolabs/talos/pull/9202

But older CLI + newer Talos - no problem.

So I don't think we need a backport (though I didn't check talosctl 1.7 + talos 1.6 combination).