Repository navigation
make lazy upgrade adhere maintenance windows - #3204
Merged
Merged
Conversation
FxKu
requested review from
Jan-M,
hughcapet,
idanovinda,
jopadi and
mikkeloscar
as code owners
October 19, 2026 15:46
…ternal pod object because in the new code we check it after calling the function
Contributor
There was a problem hiding this comment.
🟡 Changes recommended
Unresolved rotation and maintenance-window regressions can defer required restarts, strand pending upgrades, or restart a primary without a switchover.
6 open issues
What changed in this PR
This PR makes lazy Spilo upgrades use enabled maintenance windows, while preserving deferred rotation when no windows are configured.
Changes:
- Distinguishes lazy upgrades from mandatory restarts through pod annotations.
- Filters pending lazy rotations according to maintenance windows.
- Extends end-to-end coverage and documents the updated behavior.
| File | Description |
|---|---|
pkg/cluster/util.go |
Extracts maintenance-window availability checks. |
pkg/cluster/sync.go |
Filters pending rotations using lazy-upgrade policy. |
pkg/cluster/resources.go |
Defines the lazy-upgrade annotation value. |
pkg/cluster/pod.go |
Classifies rotation reasons and stores annotation values. |
pkg/cluster/connection_pooler.go |
Adapts pooler restart checks to string flags. |
pkg/cluster/cluster.go |
Includes image differences in container comparisons. |
e2e/tests/test_e2e.py |
Tests maintenance-window rotation and resets synchronous mode. |
e2e/tests/k8s_api.py |
Adds a helper to read rotation annotations. |
docs/administrator.md |
Explains maintenance-aware lazy upgrades. |
Review effort: Balanced
| func (c *Cluster) getRollingUpdateReason(cmp *compareStatefulsetResult) string { | ||
| // if the only reason is a Spilo image update and lazy update is enabled | ||
| // provide a different message so a different pod annotation is used | ||
| imageDiffOnly := len(cmp.reasons) == 1 && strings.Contains(cmp.reasons[0], "image does not match the current one") |
Contributor
Author
There was a problem hiding this comment.
@copilot we can add more checks inside getRollingUpdateReason but I would not change the current compareStatefulSet logic to not make the code more complex than it already is. Note, that image updates are managed centrally by the DBaaS team, while other updates are done by the teams. It's very unlikely that both image and other changes fall into one sync.
Updated maintenance window definitions to use fixed hours instead of dynamic time calculations. Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com>
hughcapet
approved these changes
Oct 26, 2026
Co-authored-by: Polina Bungina <27892524+hughcapet@users.noreply.github.com>
Contributor
Author
|
👍 |
hughcapet
approved these changes
Oct 27, 2026
hughcapet
approved these changes
Oct 27, 2026
FxKu
added
major
Contributor
Author
|
👍 |
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
While lazy upgrade are nice to save us from too many switchovers it becomes annoying when rolling out bugfixes in an entire fleet. Sure, we can switch it off an on, but then the rotation might not be carried out at the right time for some clusters.
Therefore, this PR proposes a solution in between. If maintenance windows are enabled and defined delay the next rolling update until that period. If no maintenance windows are configure the lazy upgrade behaves as before - waiting until the next K8s rotation (could be weeks).