Conversation
|
[APPROVALNOTIFIER] This PR is APPROVED This pull-request has been approved by: costinm The full list of commands accepted by this bot can be found here. The pull request process is described here DetailsNeeds approval from an approver in each of these files:
Approvers can indicate their approval by writing |
Codecov Report
@@ Coverage Diff @@
## master #6641 +/- ##
=======================================
+ Coverage 68% 68% +1%
=======================================
Files 361 361
Lines 31515 31701 +186
=======================================
+ Hits 21271 21442 +171
- Misses 9382 9399 +17
+ Partials 862 860 -2
Continue to review full report at Codecov.
|
|
@costinm: The following test failed, say
DetailsInstructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes/test-infra repository. I understand the commands that are listed here. |
|
|
||
| if time.Since(lastClearCache) > time.Duration(clearCacheTime)*time.Second { | ||
| log.Infof("Cleared discovery service cache after %v", time.Since(lastClearCache)) | ||
| lastClearCache = time.Now() |
There was a problem hiding this comment.
This block seems unnecessary given that the previous if block would have triggered a push on the first event in a burst of config events
There was a problem hiding this comment.
It's for cases where changes are made continuously - for example a larger system where an endpoint or config changes every 1/2 second for a longer time. We'll still do pushes every clearCacheTime. We can't do them every second because it would kill pilot (memory overhead is still high).
* Revert "Remove v2 transition commands since everything is now v2 (#6665)" This reverts commit 6339eb6. * Revert "Pilot param clusterRegistriesNamespace should default to pilot namespace (#6446)" This reverts commit b9294f7. * Revert "iptable just "return" by uid as the parameter u indicates (#6561)" This reverts commit 22a0b88. * Revert "Remove node agent service, residue from flexvolume driver. (#6651)" This reverts commit db3da82. * Revert "Continuously reapply galley CA bundle to prevent overwrite (#6599)" This reverts commit f9e8fd8. * Revert "Do not count typeConfigs if it is error. (#6527)" This reverts commit eb1de31. * Revert "Make racetest green - Fixed data races and flakiness (#6625)" This reverts commit 30b8ecb. * Revert "Improve push squashing (#6641)" This reverts commit 399cd2d.
* Improve cache push * Format
| # | ||
| galley: | ||
| enabled: true | ||
| enabled: false |
There was a problem hiding this comment.
I missed this change due to the PR title. Could you please cc @ozevren and myself in the future for these types of changes.
There was a problem hiding this comment.
Why do we have this change?
+1 w/ @ayj. This should have been properly code reviewed before merging it in.
There was a problem hiding this comment.
This PR was reverted in #6677.
FWIW, this change did expose a bug in the post-install jobs when galley was disabled. Working on a fix for now that now.
* Improve cache push * Format * Improve push squashing (#6641) * Improve cache push * Format * Revert experiment to verify if validation can be disabled
* Improve cache push * Format
* Revert "Remove v2 transition commands since everything is now v2 (istio#6665)" This reverts commit 6339eb6. * Revert "Pilot param clusterRegistriesNamespace should default to pilot namespace (istio#6446)" This reverts commit b9294f7. * Revert "iptable just "return" by uid as the parameter u indicates (istio#6561)" This reverts commit 22a0b88. * Revert "Remove node agent service, residue from flexvolume driver. (istio#6651)" This reverts commit db3da82. * Revert "Continuously reapply galley CA bundle to prevent overwrite (istio#6599)" This reverts commit f9e8fd8. * Revert "Do not count typeConfigs if it is error. (istio#6527)" This reverts commit eb1de31. * Revert "Make racetest green - Fixed data races and flakiness (istio#6625)" This reverts commit 30b8ecb. * Revert "Improve push squashing (istio#6641)" This reverts commit 399cd2d.
* Improve cache push * Format * Improve push squashing (istio#6641) * Improve cache push * Format * Revert experiment to verify if validation can be disabled
Fixes #5588 - the root problem was that events were squashed, but nothing triggered a push for the last events. So change #1 was pushed immediately, change #2 triggered a callback after (squash time).
Instead the callback is now after 1 second, and we evaluate if in the last second we received any event.
If not - push. Otherwise, keep waiting until the batch of changes settles down.