vtgate/buffer: Fix leakage of buffer pool slots due to canceled requests. #2599
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
…sts. Canceled requests, e.g. those with a deadline shorter than the failover duration, would not return their buffer pool slot. Eventually, all slots would have leaked and the buffer would start and stop buffering as usual but reject all requests immediately because it assumed there are other pending failovers which are holding the needed slots. Previously, the code path for canceled requests differed from the common method unblockAndWait() which would unblock a request and also release its buffer pool slot after the request finished its retry. This made this oversight possible. I've changed this now: canceled requests also use unblockAndWait() now. The code was also out of sync with the documentation for WaitForFailoverEnd(): The RetryDoneFunc() must not be returned when the function returns an error as well. However, for canceled requests both were returned.
alainjobart
approved these changes
Feb 26, 2017
sougou
reviewed
Feb 26, 2017
|
|
||
| // Size returns the current number of available slots. | ||
| func (sem *Semaphore) Size() int { | ||
| return len(sem.slots) |
There was a problem hiding this comment.
I think this racy, unless the language guarantees it. I couldn't find any text about this.
| // returned when the request has finished. See also shardBuffer.unblockAndWait(). | ||
| func waitForPoolSlots(b *Buffer, want int) error { | ||
| start := time.Now() | ||
| for { |
There was a problem hiding this comment.
This is a tight loop and will hang in non-preemptive implementations of the runtime. Even otherwise, it's something we should avoid.
michael-berlin
added a commit
to michael-berlin/vitess
that referenced
this pull request
Apr 19, 2017
Addressed code review comment in vitessio#2599.
frouioui
pushed a commit
to planetscale/vitess
that referenced
this pull request
Mar 26, 2024
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.
Canceled requests, e.g. those with a deadline shorter than the failover duration, would not return their buffer pool slot. Eventually, all slots would have leaked and the buffer would start and stop buffering as usual but reject all requests immediately because it assumed there are other pending failovers which are holding the needed slots.
Previously, the code path for canceled requests differed from the common method unblockAndWait() which would unblock a request and also release its buffer pool slot after the request finished its retry. This made this oversight possible. I've changed this now: canceled requests also use unblockAndWait() now.
The code was also out of sync with the documentation for WaitForFailoverEnd(): The RetryDoneFunc() must not be returned when the function returns an error as well. However, for canceled requests both were returned.