Consider ILLEGAL_GENERATION error as rebalancing error

[jakewins]

This error is indicating that the consumer is trying to commit offsets, but the consumer group has changed to a new generation.

Retrying within the existing session will indeed not work, but rejoining the group and re-trying should be successful.

Fixes #1466

[h0od]

I just encountered the error which causes my consumer to end up in crashed state (which sends shutdown signal to the process).
So as the PR suggests, a simple rejoin should be sufficient in this case.

[rpastore-wolt]

Also we are facing this error. @jakewins do you know if you PR is going to be merged or should be used a different way to handle the error ?

thanks 🙏

[kevinhering]

We're also watching this one closely. We have quite a number of kafka consumers & we recently had to create custom crash listeners to deal with just this case.

[rpastore-wolt]

We're also watching this one closely. We have quite a number of kafka consumers & we recently had to create custom crash listeners to deal with just this case.

hey @kevinhering :) thanks !
Could you share with me your solution ?

[ErlendFax]

Any reason this is not merged?

@Nevon, sorry to tag you. If you have time please consider taking a look at this and #1466

[kevinhering]

@rpastore-wolt I apologize for the delay. We've implemented a crash listener in our consuming code to listen for this particular event (see the docs on Instrumenting Events for some info about event listeners & the crash event). Here's the basic idea:

consumer.on(CRASH, (event) => handleCrashEvent(event)) // listen for crash event

in the crash event handler, look for an event with payload similar to this (pay particular attention to values in error.retriable & error.cause.type):

{
  "topic": "topic-name",
  "error": {
    "name": "KafkaJSNonRetriableError",
    "retriable": false,
    "cause": {
      "name": "KafkaJSProtocolError",
      "retriable": false,
      "type": "ILLEGAL_GENERATION",
      "code": 22
    }
  }
}

You can add whatever code you need in order to alert/restart the consumer from here.

Hope that helps. (and hope we won't need this soon!)

[patrykwegrzyn]

ILLEGAL_GENERATION has become a pain for us recently , any idea will this be merged anytime soon or if is going to be merged at all?

[robinbijlani]

Also been watching this PR for months, very eager to see it hopefully merged in. 🤞 🤞

[jgoldsmith613]

We are waiting on this too. Any timeline on a a fix?
This is straight from confluent support
"To reduce impact for both cases above, the client will need to be designed to retry its connection to Kafka if the ILLEGAL_GENERATION exception is encountered. "

[josh-cain]

We're hitting this too, and it's causing a bit of toil. Would really appreciate a merge 🙏 !

[ecalvert]

We're hitting this as well. Causing major issues when confluent cloud performs their cluster rolls. We see consumer crashes every 5-10 minutes for a 2 hour span

[srMarquinho]

Just checking this issue is active and being looked after. This is exactly our problem and it is absolutely major to us. Thank you all. 🙌

[Nevon]

This bug has been resolved in v2.2.4. Thanks @jakewins for the fix!