scheduler: perform feasibility checks for system canaries before computing placements #26953
+201
−417
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
…uting placements Canaries for system jobs are placed on a tg.update.canary percent of eligible nodes. Some of these nodes may not be feasible, and until now we removed infeasible nodes during placement computation. However, if it happens to be that the first eligible node we picked to place a canary on is infeasible, this will lead to the scheduler halting deployment. The solution presented here simplifies canary deployments: initially, system jobs that use canary updates get allocations placed on all eligible nodes, but before we start computing actual placements, a method called `evictCanaries` is called (much like `evictAndPlace` is for honoring MaxParallel), and performs a feasibility check on each node up to the amount of required canaries per task group. Feasibility checks are expensive, but this way we only check all the nodes in the worst case scenario (with canary=100), otherwise we stop checks once we know we are ready to place enough canaries.
tgross
reviewed
Oct 17, 2025
Comment on lines
+608
to
+613
| // we only now the total amountn of placements once we filter out | ||
| // infeasible nodes, so for system jobs we do it backwards a bit: the | ||
| // "desired" total is the total we were able to place. | ||
| if s.deployment != nil { | ||
| s.deployment.TaskGroups[tgName].DesiredTotal += 1 | ||
| } |
There was a problem hiding this comment.
For system jobs I think we need to make sure we're working from a blank slate dstate for each evaluation. Incrementing this here is adding on top of the desired total from previous evals.
Comment on lines
+761
to
+766
| // ensure everything is healthy | ||
| if dstate, ok := s.deployment.TaskGroups[groupName]; ok { | ||
| if dstate.HealthyAllocs < dstate.DesiredTotal { // Make sure we have enough healthy allocs | ||
| complete = false | ||
| } | ||
| } |
There was a problem hiding this comment.
If we're resetting desired total in computePlacements, it won't be correctly set when we reach isDeploymentComplete, which is called before that.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Canaries for system jobs are placed on a tg.update.canary percent of eligible nodes. Some of these nodes may not be feasible, and until now we removed infeasible nodes during placement computation. However, if it happens to be that the first eligible node we picked to place a canary on is infeasible, this will lead to the scheduler halting deployment.
The solution presented here simplifies canary deployments: initially, system jobs that use canary updates get allocations placed on all eligible nodes, but before we start computing actual placements, a method called
evictCanariesis called (much likeevictAndPlaceis for honoring MaxParallel), and performs a feasibility check on each node up to the amount of required canaries per task group. Feasibility checks are expensive, but this way we only check all the nodes in the worst case scenario (with canary=100), otherwise we stop checks once we know we are ready to place enough canaries.