Closed mavdi closed 6 years ago
Hi @mavdi
at a first glance, this seems to be an issue with monitoring interceptors and not with the S3 connector itself.
And some additional information:
flush.size
without interceptors. debug
or trace
, focus on specific loggers by keeping some irrelevant loggers to info
or debug
levels).I've removed the interceptors and it seems to have allowed me to increase flush.size
to 40000
.
Anything beyond that still fails. I get a strange error:
WARN Commit of WorkerSinkTask{id=kafka-s3-sink-time20-0} offsets timed out
Again, hit and miss kind of errors. I can't make any concise picture of what's going on..
[2017-10-18 09:29:18,749] WARN Aborting multi-part upload with id 'z5wjp1k4gZvSEHiGAh3OQC_bhI9CUOybf8x._xvuWtOqOOkgc8gVeX1h3AEUrlgjWLQHe9.GXAI9OKrx3jwV72xNgV9vzGXu0R83IughQLs9gdbWU19BTQatOg9QEsUazu_V1H8B37ncZtKNhyMgp77wk1TJ2dWGYUUmIDdFByo-' (io.confluent.connect.s3.storage.S3OutputStream)
A higher flush.size
is definitely attainable by the connector. Aborts of multi-part uploads might be more related to connectivity issues. Tuning s3.part.size
could help diagnose such issue for different sizes. Closing because it's been a while since this issue was discussed, but feel free to re-open if it remains an issue.
I'm setting up a Kafka connect S3 sink with the following configuration
When setting
flush.size
to about10000
everything works. But any number greater than that, seems to hang the connector altogether. I'm saying this because I get the following error:Is this a known issue? Any workarounds?