sql, pgwire: cancel sessions better #30813
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
... and comments around the SessionRegistry. Release note: None
Before this patch, `cancel session <foo>` would cause a connExecutor's context to be canceled. This patch switches it to cancel the pgwire connection's ctx instead (which was a parent of the connExecutor's ctx; now the two are the same), as nature intended. For example, reading from the network connection should stop ASAP after `cancel session`. It also opens to door to guaranteeing that the client using the respective session is unblocked in a timely manner (by the server closing the network connection), although that is not currently implemented (pgwire will still wait for the connExecutor to react to cancelation and finish its task before closing the net.Conn). This patch also eliminates what I presume to have been a crash when someone tries to cancel a "session" created by the InternalExecutor. It will now be a no-op. This patch also moves a cancelation callback out of connExecutor.ctxHolder. I'm trying to get rid of that guy, which is how this patch came to be. I've also added a TODO with some thoughts on a better design for session cancelation, originally expressed in cockroachdb#23861 (comment). FWIW, I've tried to be the change, but gave up for the moment. Release note: None
|
This change is |
knz
approved these changes
Oct 1, 2018
There was a problem hiding this comment.
Reviewed 2 of 2 files at r1, 2 of 2 files at r2.
Reviewable status:complete! 0 of 0 LGTMs obtained (and 1 stale)
andreimatei
commented
Oct 1, 2018
Build failed |
andreimatei
commented
Oct 1, 2018
craig bot
pushed a commit
that referenced
this pull request
Oct 1, 2018
30813: sql, pgwire: cancel sessions better r=andreimatei a=andreimatei Before this patch, `cancel session <foo>` would cause a connExecutor's context to be canceled. This patch switches it to cancel the pgwire connection's ctx instead (which was a parent of the connExecutor's ctx; now the two are the same), as nature intended. For example, reading from the network connection should stop ASAP after `cancel session`. It also opens to door to guaranteeing that the client using the respective session is unblocked in a timely manner (by the server closing the network connection), although that is not currently implemented (pgwire will still wait for the connExecutor to react to cancelation and finish its task before closing the net.Conn). This patch also eliminates what I presume to have been a crash when someone tries to cancel a "session" created by the InternalExecutor. It will now be a no-op. This patch also moves a cancelation callback out of connExecutor.ctxHolder. I'm trying to get rid of that guy, which is how this patch came to be. I've also added a TODO with some thoughts on a better design for session cancelation, originally expressed in #23861 (comment). FWIW, I've tried to be the change, but gave up for the moment. Release note: None Co-authored-by: Andrei Matei <[email protected]>
Build succeeded |
|
I would recommend a backport for this. |
|
(To go together with #31002) |
|
You sure you want to backport 31002? It seems like a pretty large change...
But if you do and you need this one for it, feel free to lump it in as far
as I'm concerned.
…On Fri, Oct 5, 2018, 6:51 AM kena ***@***.***> wrote:
(To go together with #31002
<#31002>)
—
You are receiving this because you authored the thread.
Reply to this email directly, view it on GitHub
<#30813 (comment)>,
or mute the thread
<https://github.com/notifications/unsubscribe-auth/AAXBcf9TIftYtc2twYckOrTntFCPR8PFks5uhzmXgaJpZM4XBXbp>
.
|
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.
Before this patch,
cancel session <foo>would cause a connExecutor'scontext to be canceled. This patch switches it to cancel the pgwire
connection's ctx instead (which was a parent of the connExecutor's ctx;
now the two are the same), as nature intended. For example, reading from
the network connection should stop ASAP after
cancel session. It alsoopens to door to guaranteeing that the client using the respective
session is unblocked in a timely manner (by the server closing the
network connection), although that is not currently implemented (pgwire
will still wait for the connExecutor to react to cancelation and finish
its task before closing the net.Conn).
This patch also eliminates what I presume to have been a crash when
someone tries to cancel a "session" created by the InternalExecutor. It
will now be a no-op.
This patch also moves a cancelation callback out of
connExecutor.ctxHolder. I'm trying to get rid of that guy, which is how
this patch came to be.
I've also added a TODO with some thoughts on a better design for session
cancelation, originally expressed in
#23861 (comment).
FWIW, I've tried to be the change, but gave up for the moment.
Release note: None