colexec: add more redundancy to releasing disk resources #81562
Merged
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
|
This change is |
As a couple of recently-found issues showed, making sure that all disk resources are released can be tricky since disk-backed operators can form large graphs with multiple external operators supporting a single operation. This commit makes the release of disk resources more bullet-proof by auditing all users of the vectorized disk queues to make sure they are added to `OpWithMetaInfo.ToClose` which are closed on the flow cleanup. Since `Close` can be safely called multiple times, it adds some redundancy, leaning on the side of caution. In particular, the following changes are made: - external distinct and external hash aggregators are explicitly added to `ToClose` slice. They should already be now closed by the `diskSpillerBase`, but it doesn't hurt closing them explicitly. - window aggregator operator has been refactored so that it doesn't throw an error in its `Close` method - with the previous version it was possible to panic during the `Close` execution and possibly leak some resources. - signatures of the constructor methods have been adjusted to return `ClosableOperator` to make the need for closing be more explicit. - each router output is now a `Closer` and the consumer of each output is now resposible for closing it. Again, I'm pretty sure that each output will have been closed by that time the consumer explicitly tries to close the output, yet there is no harm in closing it twice. An additional minor cleanup is the removal of the usage of an embedded context in a couple `Close` implementations given that the function takes it as an argument. Release note: None
cucaroach
approved these changes
May 20, 2022
There was a problem hiding this comment.
Reviewed 28 of 28 files at r1, all commit messages.
Reviewable status:complete! 0 of 0 LGTMs obtained (waiting on @michae2)
michae2
approved these changes
May 20, 2022
There was a problem hiding this comment.
Reviewed 21 of 28 files at r1, all commit messages.
Reviewable status:
|
TFTRs! bors r+ |
|
Build succeeded: |
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.
As a couple of recently-found issues showed, making sure that all disk
resources are released can be tricky since disk-backed operators can
form large graphs with multiple external operators supporting a single
operation. This commit makes the release of disk resources more
bullet-proof by auditing all users of the vectorized disk queues to make
sure they are added to
OpWithMetaInfo.ToClosewhich are closed on theflow cleanup. Since
Closecan be safely called multiple times, it addssome redundancy, leaning on the side of caution.
In particular, the following changes are made:
to
ToCloseslice. They should already be now closed by thediskSpillerBase, but it doesn't hurt closing them explicitly.throw an error in its
Closemethod - with the previous version it waspossible to panic during the
Closeexecution and possibly leak someresources.
ClosableOperatorto make the need for closing be more explicit.Closerand the consumer of each outputis now resposible for closing it. Again, I'm pretty sure that each
output will have been closed by that time the consumer explicitly tries
to close the output, yet there is no harm in closing it twice.
An additional minor cleanup is the removal of the usage of an embedded
context in a couple
Closeimplementations given that the functiontakes it as an argument.
Release note: None