DESCRIPTION

CBQE:0 Sometimes replicators might take longer to sleep after previous operation so add some buffer time before loading any mutations

Change-Id: Ie2c6599d3257a9fe53ae63e7b49a04cbbbb4fea1


COMMENTS

author: Hugo Blankenship
date: 2014-07-25 22:55:53.368000000

Patch Set 2:

Build Started http://factory.couchbase.com/job/testrunner-Jasmin Rangel-master/2150/

-------------------------------------
author: Angelina Mayer
date: 2014-07-25 22:58:33.002000000

Patch Set 2:

These tests fail on Jenkins but not when run locally. On analyzing logs, it was found that there was not much time gap between the replicator going to sleep and waking up for the next mutation.

-------------------------------------
author: Angelina Mayer
date: 2014-07-25 22:58:41.075000000

Patch Set 2: Verified+1

-------------------------------------
author: Hugo Blankenship
date: 2014-07-25 23:15:31.162000000

Patch Set 2: Verified+1

Build Successful 

http://factory.couchbase.com/job/testrunner-Jasmin Rangel-master/2150/ : SUCCESS

-------------------------------------
author: Esperanza Cherry
date: 2014-07-28 08:47:07.718000000

Patch Set 2:

(1 comment)

Line:324, pytests/xdcr/checkpointXDCR.py -> it's better don't use hard code sleep time, self.wait_timeout is an option for such cases. and please add the reason why we do this sleep(ticket number or description)

-------------------------------------
author: Angelina Mayer
date: 2014-07-28 22:33:01.064000000

Patch Set 2:

(1 comment)

Line:324, pytests/xdcr/checkpointXDCR.py -> We added sleep for this reason I put as the commit msg - "Sometimes replicators might take longer to sleep after previous operation so add some buffer time before loading any mutations"

wait_timeout would make sense if we know for sure we're waiting for the server to complete some operation. But when a replicator goes to sleep after finishing its work is not exactly deterministic. When I run the same test on my environment, it passes, sometimes fails and the logs in the latter case show that there wasn't enough time gap between replicator sleep time and wake-up time to expect a checkpoint. And there's nothing to timeout on. This is just buffer time to make sure checkpointing happens.

So would prefer to keep just the sleep. Copied the same comment from below in next changeset.

-------------------------------------
author: Esperanza Cherry
date: 2014-07-30 06:25:56.895000000

Patch Set 2:

(1 comment)

forgot to commit my comments yesterday

Line:324, pytests/xdcr/checkpointXDCR.py -> one example:
for example on build N 100 sec is not enought again. what we can do: try to set 150, 200 ... change the code and rerun the tests. another option is to increase it in tests params -t wait_timeoit=100 and run on slow envs; on your fast vms/localy you don't need to wait extra time at the same time.

-------------------------------------
