Conversation
This test triggers heartbeat failrues which leads to REBALANCE_IN_PROGRESS which causes hang.
blinkov
added
the
pr-not-for-changelog
|
Thank @azat, I have some theory about how to workaround that |
Great! FWIW I was thinking about adding |
|
My first attempts to use the test dont hang... :/ i've played a bit with timeout, but no success yet |
|
Hm, can you share logs (stderr, clickhouse*.log), so I can take a look (or/and compare with mine)? |
|
|
||
| std::atomic<bool> stub = {false}; | ||
| copyData(*in, *block_io.out, &stub); | ||
| copyData(*in, *block_io.out, &stream_cancelled); |
There was a problem hiding this comment.
missed that. We can't cancel during copyData, as it's not aware of commits and other kafka-related stuff.
It will be cancelled on underlying layer (kafka buffer)
There was a problem hiding this comment.
Yep, I already thought about this, but forgot, thanks!
There was a problem hiding this comment.
But better to add some comment, to avoid further errors
Kafka cleanup (cherry picked from commit 4df7dd8)
Kafka cleanup (cherry picked from commit 4df7dd8)
5 participants
Changelog category (leave one):
This PR contains:
[thrd:main]: Group "consumer_hang": waiting for rebalance_cb, 0 toppar(s), 0 unassignment(s), 0 commit(s) (state up, join-state wait-assign-rebalance_cb) before terminatingmessages in stderr)kafka_max_block_sizeandkafka_commit_every_batch)Cc: @filimonov
Cc: @abyss7
Cc: @vavrusa