You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Basically we want to be able to update any component in rolling fashion over built-in k8s features.
There are roughly 3 parts:
[x] Have an API to be able to set update strategy on per-component basis, which was implemented in #383 — now we just need to add strategy field to update plan
[ ] Current implementation of scale-to-zero update will stay to cover some cases when rolling is not possible, but it need to be generalized/refactored, because now it basically contains 1) scale to zero/remove pods step, 2) update sts 3) wait pods created — which should be kinda squashed in one step in the update flow, because rolling wouldn't have remove all pods/create all pods stages. Unfortunately logic of handling component update is scatterred/partly copy-pasted in all (almost 20) components, so it would take time. This should be done as a separate PR before implementing rolling.
[ ] The rolling itself instead of scaling to zero, should update required fields (in 99% it is an image or config), set update window and track sts progress on update. Details are here #284
reacted with thumbs up emoji reacted with thumbs down emoji reacted with laugh emoji reacted with hooray emoji reacted with confused emoji reacted with heart emoji reacted with rocket emoji reacted with eyes emoji
-
Basically we want to be able to update any component in rolling fashion over built-in k8s features.
There are roughly 3 parts:
[x] Have an API to be able to set update strategy on per-component basis, which was implemented in #383 — now we just need to add strategy field to update plan
[ ] Current implementation of scale-to-zero update will stay to cover some cases when rolling is not possible, but it need to be generalized/refactored, because now it basically contains 1) scale to zero/remove pods step, 2) update sts 3) wait pods created — which should be kinda squashed in one step in the update flow, because rolling wouldn't have remove all pods/create all pods stages. Unfortunately logic of handling component update is scatterred/partly copy-pasted in all (almost 20) components, so it would take time. This should be done as a separate PR before implementing rolling.
[ ] The rolling itself instead of scaling to zero, should update required fields (in 99% it is an image or config), set update window and track sts progress on update. Details are here #284
Beta Was this translation helpful? Give feedback.
All reactions