Remove lock from sessionWS.SendBytes #1140
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
There are six different calls to s.Lock() in sessionWS. Three of these locks are around calls to SetWriteDeadline and WriteMessage on the WebSocket. WriteMessage will block for up to the timeout set in SetWriteDeadline, which has a default of 5000 milliseconds. This lock is also acquired by SendBytes to determine if the session has been stopped. This is a problem as the event processing methods in StatusRegistry and Track call SendBytes directly on the Session. If one of the Sessions being written to has delays in writing data, this could cause backups in the event processing. The simplest approach (this commit) is to just remove the lock within SendBytes. To avoid a race condition with writing to a closed channel, we need to not close the outgoingCh channel in the Close method. This is safe to do as channels do not need to be explicitly closed and will automatically be cleaned up by the garbage collector. It should also be possible to remove the lock entirely from sessionWS, but that is a much larger change.
|
|
joshgarnett
added a commit
to joshgarnett/nakama
that referenced
this pull request
Nov 14, 2023
There are six different calls to s.Lock() in sessionWS. Three of these locks are around calls to SetWriteDeadline and WriteMessage on the WebSocket. WriteMessage will block for up to the timeout set in SetWriteDeadline, which has a default of 5000 milliseconds. This lock is also acquired by SendBytes to determine if the session has been stopped. This is a problem as the event processing methods in StatusRegistry and Track call SendBytes directly on the Session. If one of the Sessions being written to has delays in writing data, this could cause backups in the event processing. This change is more complex than heroiclabs#1140. It fully removes the mutex from sessionWS. All writes have been moved to only occur within processOutgoing so no panics are generated by concurrent writes. The stopped variable is converted to an atomic.Bool, which is only used in the Close method to prevent more than one Close call being made. As the outgoingCh is not closed, Send() is used to send the final messages and then the ctxCancelFn is called, which causes a Close message to be sent in processOutgoing after all messages are sent. The ping backoff logic was also simplified.
zyro
approved these changes
Jan 24, 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.
There are six different calls to s.Lock() in sessionWS. Three of these locks are around calls to SetWriteDeadline and WriteMessage on the WebSocket. WriteMessage will block for up to the timeout set in SetWriteDeadline, which has a default of 5000 milliseconds. This lock is also acquired by SendBytes to determine if the session has been stopped. This is a problem as the event processing methods in StatusRegistry and Track call SendBytes directly on the Session. If one of the Sessions being written to has delays in writing data, this could cause backups in the event processing.
The simplest approach (this commit) is to just remove the lock within SendBytes. To avoid a race condition with writing to a closed channel, we need to not close the outgoingCh channel in the Close method. This is safe to do as channels do not need to be explicitly closed and will automatically be cleaned up by the garbage collector. It should also be possible to remove the lock entirely from sessionWS, but that is a much larger change.