Details
-
Bug
-
Resolution: Fixed
-
Critical
-
3.0
-
Security Level: Public
-
None
-
Untriaged
-
Unknown
-
June 30 - July 18
Description
When doing a stream request with the on-disk items only flag (0x02) set, ep-engine does something strange that leads to a timeout in the UPR client.
Please note that I know little about ep-engine, but here's what I found out. I've attached to log files. The no-flag.log is without the flag set, the on-disk-only-flag.log with the flag set.
You can see in the on-disk-only-flag.log that a stream for vBuclet 0 is created twice:
memcached<0.86.0>: Tue Apr 22 15:50:29.918776 CEST 3: (beer-sample) UPR (Producer) eq_uprq:Indexer mapreduce_view: beer-sample _design/beer (prod/main) - Stream created for vbucket 0
memcached<0.86.0>: Tue Apr 22 15:50:38.292332 CEST 3: (beer-sample) UPR (Producer) eq_uprq:Indexer mapreduce_view: beer-sample _design/beer (prod/main) - Stream created for vbucket 0
It seems that the UPR client receives the snapshot information already with the first stream request. It then waits for more data. The current timeout is set to 5 seconds, which is not enough, as the second stream, that seems to contain the mutations starts 8 seconds later.