Fix for transactions not allowed to finish during PlannedReparentShard #8098
Conversation
This file contains 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
Backport of vitessio#8089 This is a combination of 3 commits. * remove precheck of tablet serving and target * remove the additional logic and return error if queryservice not found to serve query * fix test as per new change Signed-off-by: Harshit Gangal <harshit@planetscale.com> Signed-off-by: Andres Taylor <andres@planetscale.com>
|
some unit tests are failing. |
|
As discussed offline, let us create an issue the describes the regression and link to the backport PRs. |
deepthi
previously approved these changes
May 13, 2021
Signed-off-by: deepthi <deepthi@planetscale.com>
We should have listened to @harshit-gangal. One of the perils of doing backports is that we may pull in changes that don't make sense on an older branch. That is what seems to have happened here. Tests should pass with the latest commit that reverts the unwanted changes. |
systay
changed the title
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Outdated suggestions cannot be applied.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.
An issue was introduced In a recent fix (#8041) which was a backport of #7879
Before sending queries to a tablet, #7879 changed the behaviour to check if the tablet is ready to answer, by checking it's ServingStatus and that the tablet type hasn't changed.
If PlannedReparentShard is going on, this check should not be done for transactions in flight.
The vttablet waits for inflight transactions to get commit/rollback i.e. we want queries to existing transaction to be sent down to get the transaction completed, even if the tablet is currently saying it is NotServing.
The test that exposed this issue was already in the code base: go/test/endtoend/tabletgateway/buffer/buffer_test.go became flaky after #7879 was merged.
So, to fix the issue, the pre-check is removed from Gateway when getting the tablet connection for existing active shard_sessions with vttablet.
This also means that the reserved connection that used to reset based on this pre-check logic will have to hit the vttablet first and then only will reset the shard session on receiving the expected error making it two round trips.
Related Issue(s)
Backport of #8089