sdk: Client ping on idle connections #2309
Open
+75
−6
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
kim
commented
Feb 26, 2025
| "Error reading message from read WebSocket stream", | ||
| Err(e), | ||
| ); | ||
| break; |
There was a problem hiding this comment.
None of the other error variants indicate a particularly recoverable situation, so I added this break.
Make the websocket loop send a `Ping` frame when the connection is idle for a while. "idle" includes missing server-side pings. Writing to the socket will detect if the remote end went away, but FIN / RST went missing for some reason. This is not detected by only reading.
kim
force-pushed
the
kim/sdk/client-ping
branch
from
081ce29 to
abda910
Compare
gefjon
reviewed
Feb 26, 2025
| record_metrics(other.len()); | ||
| }, | ||
| }, | ||
|
|
||
| _ = idle_timeout_interval.tick() => { | ||
| if mem::replace(&mut idle, true) { | ||
| if mem::take(&mut want_pong) { |
There was a problem hiding this comment.
Why are we mem::takeing here? It looks to me like this is equivalent to just reading want_pong in all cases.
want_pong == true:takeand replace withfalse. Break out of the loop and never read fromwant_pongagain.want_pong == false:takeis a no-op. Setwant_pongtotruelater in this block. Never read thefalsevalue installed bymem::take.
It seems like there are two possible simplifications to this code:
mem::replace(&mut want_pong, true)here, and remove thewant_pong = trueassignment on line 390.- Just branch on
want_pongwithif want_pong {here, and leave the assignment later on.
Comment on lines
+386
to
+390
| Self::maybe_log_error( | ||
| "Error sending ping", | ||
| self.sock.send(WebSocketMessage::Ping(vec![])).await, | ||
| ); | ||
| want_pong = true; |
There was a problem hiding this comment.
Do we want to break the loop if we fail to send a ping?
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.
Make the websocket loop send a
Pingframe when the connection is idle for a while. "idle" includes missing server-side pings.Writing to the socket will detect if the remote end went away, but FIN / RST went missing for some reason. This is not detected by only reading.
I haven't touched the scary server-side websocket loop, but we may consider to also make it send pings only if it doesn't have anything else to send.
Expected complexity level and risk
2
Testing
The situation can be approximated by inducing a 100% packet loss scenario.
This can be done in various ways. A convenient one is
docker compose disconnect, which disables the virtual interface of a container.Without this patch, you'll see the server timing out the connection after a while (as it did not receive pong), while the client keeps thinking that something may eventually arrive.
With this patch, both ends eventually close the connection.