Open PeterGrace opened 1 year ago
I personally would go with the first option. In my understaing, taints are a way to let only specific pods to be allowed on specific node, which is also aim of this project I suppose. If using affinities, my understating would be more about distribution of pods across different nodes, in this case distribution between different architectures. If I'm mistaken, please put me in the correct direction.
I personally would go with the first option. In my understaing, taints are a way to let only specific pods to be allowed on specific node, which is also aim of this project I suppose. If using affinities, my understating would be more about distribution of pods across different nodes, in this case distribution between different architectures. If I'm mistaken, please put me in the correct direction.
Thanks for the feedback! Yes, I do think that the initial scan of daemonsets is probably the right idea. I do need to revisit this since I hit this issue this morning during upgrades of a daemonset and remembered I should solve this problem. 🤣
Unresolved issue caught you up at the right moment 😄 . If you need any help I might be at your disposal. Although I never coded in Rust before.
Thanks for the feedback! Yes, I do think that the initial scan of daemonsets is probably the right idea. I do need to revisit this since I hit this issue this morning during upgrades of a daemonset and remembered I should solve this problem. 🤣
Narrator: he hit this issue this morning as well :rofl:
DaemonSets won't schedule on tainted nodes unless they tolerate them. Because tolerable is a MutatingWebhookController, it only operates on items being created or modified. This means that when tolerable is installed in a running cluster, any pre-existing DaemonSets wouldn't be edited to allow the toleration.
How do I solve this? Ideas: