Tim Armstrong has posted comments on this change.

Change subject: Test for IMPALA-2987
......................................................................


Patch Set 2:

(1 comment)

http://gerrit.cloudera.org:8080/#/c/2306/2/tests/custom_cluster/test_exchange_delays.py
File tests/custom_cluster/test_exchange_delays.py:

Line 28: 
@CustomClusterTestSuite.with_args("--stress_data_stream_recvr_delay_ms=1000"
       :         " --datastream_timeout_ms=500")
       :   def test_exchange_small_delay(self, vector):
       :     """Test delays in registering data stream receivers where the 
receiver registers
       :     relatively soon after the timeout.
       :     """
       :     self.run_test_case('QueryTest/exchange-delays', vector)
       : 
       :   @pytest.mark.execute_serially
       :   
@CustomClusterTestSuite.with_args("--stress_data_stream_recvr_delay_ms=10000"
       :         " --datastream_timeout_ms=1")
> What's the thought behind testing different delays? I worry that in the sho
It shouldn't be that noisy, it's a custom cluster test that runs serially so 
the query is the only thing running.

The two delays test different failure modes: one resulted in incorrect results 
and the other a hang.

In the short delay case, the first batch and maybe second batch time out, but 
later batches will arrive once the receiver is present. This gave incorrect 
results before your fix. I guess it's possible that the sender is delayed by 
over 500ms so there's no timeout, but it seems pretty unlikely.

In the long delay case, all of the batches will typically arrive before the 
timeout. This resulted in a hang before your fix.


-- 
To view, visit http://gerrit.cloudera.org:8080/2306
To unsubscribe, visit http://gerrit.cloudera.org:8080/settings

Gerrit-MessageType: comment
Gerrit-Change-Id: Id5f5746b6023c301bacfa305c525846cdde822c9
Gerrit-PatchSet: 2
Gerrit-Project: Impala
Gerrit-Branch: cdh5-2.5.0_5.7.0
Gerrit-Owner: Tim Armstrong <[email protected]>
Gerrit-Reviewer: Dan Hecht <[email protected]>
Gerrit-Reviewer: Henry Robinson <[email protected]>
Gerrit-Reviewer: Michael Ho
Gerrit-Reviewer: Tim Armstrong <[email protected]>
Gerrit-HasComments: Yes

Reply via email to