The logs show, that the we process about 25,000 messages. Thus, the question 
is, if we can assume that we hit the "insert case" with high enough 
probability? Key space might still be too small? Thoughts?

Note, that we don't want to "exhaust" the bigger key space -- if we exhaust it 
key space, it implies that we used all keys before we crash an application, and 
thus, only test update case (which worked correct even before the fix). We want 
to test the "insert new key into store" case, and make sure that the insert is 
"reverted" (this corner case was the root cause of the bug).

Ie, with 100 keys as originally, we processed about 7500 messages before we do 
the first crash -- thus, with very high probability, we used each key already 
and thus could not expose the bug.

[ Full content available at: https://github.com/apache/kafka/pull/5640 ]
This message was relayed via gitbox.apache.org for [email protected]

Reply via email to