Make dead brokers die harder #548
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
When a broker gets an error trying to receive a response (either from the network layer, or from failing to parse the minimal global header), it should just abandon ship and die. Save that error and return it immediately for any further requests we might have made. - The vast majority of the time the connection is going to be hosed anyways, if nothing else by being out-of-sync on correlation IDs (which we don't handle and which doesn't seem particularly urgent). - All of Sarama's built-in callers (producer/consumer/offset-manager) immediately `Close` a broker when they receive one of these errors anyways, so all this does is speed up that in the common case. *If* one of these errors is recoverable, and *if* there is user-space code somewhere which actually tries to recover in one of those cases, then that code would break. This neatly satisfies one of the XXX comments I left in about this issue from way back in 2013. The TODOs about correlation ID matching are still present.
|
This change looks good to me 👍. I can't really see any use code out there trying to salvage the connection, especially because the correlation IDs are probably broken after something like this happens. |
|
Alternatively, we could put in a circuit breaker. Not sure if it is worth the effort. |
|
A circuit-breaker would be neat, but we already have other ones higher in the stack so we'd end up with nested breakers on slightly different conditions which would be odd. Also, this is really not something that can heal without re-starting the network connection so it's not an ideal use. |
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.
When a broker gets an error trying to receive a response (either from the
network layer, or from failing to parse the minimal global header), it should
just abandon ship and die. Save that error and return it immediately for any
further requests we might have made.
nothing else by being out-of-sync on correlation IDs (which we don't handle
and which doesn't seem particularly urgent).
immediately
Closea broker when they receive one of these errors anyways, soall this does is speed up that in the common case.
If one of these errors is recoverable, and if there is user-space code
somewhere which actually tries to recover in one of those cases, then that code
would break.
This neatly satisfies one of the XXX comments I left in about this issue from
way back in 2013. The TODOs about correlation ID matching are still present.
@wvanbergen @4578395263256 another approach to #546, one I am generally happier with. Thoughts?