-
Notifications
You must be signed in to change notification settings - Fork 2
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Cooperative stickies w/ metrics #43
base: qox-stats-callback
Are you sure you want to change the base?
Conversation
cd89e41
to
4455a57
Compare
|> map (\_ -> ()) | ||
flushProducer :: Producer.KafkaProducer -> Prelude.IO () | ||
flushProducer producer = do | ||
Producer.flushProducer producer |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
We've used flushProducer
before in order to use sync sending messages. I think it's also a blocking function and I think to remember that it also slowed down sending immensely. ℹ️
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
maybe it's fixed though.
We're gettingt this error: All partition.assignment.strategy (range,cooperative-sticky) assignors must have the same protocol type, online migration between assignors with different protocol types is not supported Some minimal reading of rdkafka made us believe range is EAGER and cooperative-sticky is .. well .. COOPERATIVE, which are different protocol types. This means we can't use both at the same time and in practice, means we need downtime on workers when deploying this change. Not terrible tbh.
Extra props doesn't set things up correctly to use incremental_assign and friends.
We've switched to the only remaining pollEvent proxy: flushProducer It might be fine to flush 'em every 100ms. We'll see.
0aff8f2
to
da82ac6
Compare
Try out our fork of hw-kafka allowing cooperative rebalance strategy, based off of our WIP PR exposing rdkafka metrics, so we can understand what's up with rdkafka under the hood.