Pulsar: [Reader] seek to batch messageId went wrong in 2.5.0

Created on 17 Feb 2020  路  4Comments  路  Source: apache/pulsar

Starting the reader with a MessageId inside a batch: (11:0:3:151) with :

client.newReader()
         .topic(topic)
         .startMessageId(startMessageId)
         .startMessageIdInclusive()
         .loadConf(readerConf)
         .create();

the first message I read with reader.readNext() happens to be the one with messageId (11:0:3:0).

This is 2.5.0 related since my library relies on 2.4.2 works correctly.

triagweek-8 typbug

Most helpful comment

Hi @lhotari @devinbost , I found this bug is brought in by #5622 , and I fixes it in #6345 . I'll look into #6333 and #6224 now.

All 4 comments

I wonder if #6333 is related.

@yjshen Do you think this one might also be related?
If we can already reproduce this in a test, then we have a clear path to fixing it.
https://github.com/apache/pulsar/issues/6224

Hi @lhotari @devinbost , I found this bug is brought in by #5622 , and I fixes it in #6345 . I'll look into #6333 and #6224 now.

6224 is a bug but not related to this issue. @devinbost will fix it in his PR.

Was this page helpful?
0 / 5 - 0 ratings