Zhongwei Zhu created SPARK-32044:
------------------------------------
Summary: Kakfa continuous processing print mislead initial offsets
log
Key: SPARK-32044
URL: https://issues.apache.org/jira/browse/SPARK-32044
Project: Spark
Issue Type: Bug
Components: Structured Streaming
Affects Versions: 2.4.6
Reporter: Zhongwei Zhu
Fix For: 2.4.7
When using structured streaming in continuous processing mode, after restart
spark job, spark job can correctly pick up offsets in checkpoint location from
last epoch. But it always print out below log:
20/06/12 00:58:09 INFO [stream execution thread for [id =
34e5b909-f9fe-422a-89c0-081251a68693, runId =
0246e19d-aaa1-4a5c-9091-bab1a0578a0a]] kafka010.KafkaContinuousReader: Initial
offsets:
\{"kafka_topic":{"8":51618236,"11":51610655,"2":51622889,"5":51637171,"14":51637346,"13":51627784,"4":51606960,"7":51632475,"1":51636129,"10":51632212,"9":51634107,"3":51611013,"12":51626567,"15":51640774,"6":51637823,"0":51629106}}
And this log is misleading, as spark didn't use this one as initial offsets.
This is caused by below code in KafkaContinuousReader
{code:java}
offset = start.orElse {
val offsets = initialOffsets match {
case EarliestOffsetRangeLimit =>
KafkaSourceOffset(offsetReader.fetchEarliestOffsets())
case LatestOffsetRangeLimit =>
KafkaSourceOffset(offsetReader.fetchLatestOffsets(None))
case SpecificOffsetRangeLimit(p) =>
offsetReader.fetchSpecificOffsets(p, reportDataLoss)
}
logInfo(s"Initial offsets: $offsets")
offsets
}
{code}
The code inside orElse block is always executed even when start has value.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
