Hello Alexey Serbin,
I'd like you to do a code review. Please visit
http://gerrit.cloudera.org:8080/11523
to review the following change.
Change subject: create-table-stress-test: deflake
TestConcurrentCreateTableAndReloadMetadata
......................................................................
create-table-stress-test: deflake TestConcurrentCreateTableAndReloadMetadata
First, if the ASSERT_OK fired, the test would hang. No bueno.
Second, in TSAN environments the test would sometimes fail with a table
already present error, typically happening either 30s or 60s after that
table was created. I believe this is due to a client-side timeout _after_
the table was created but _before_ its RPC response was sent. We could
try to address this by bumping the client's timeout, but because
reload_metadata_thread is DoS'ing the cluster, that doesn't seem robust.
Instead, let's have the thread ease up in TSAN environments.
Without the change, the test failed 4/1000 times (--stress_cpu_threads=4).
With the change, the test failed 0/1000 times (--stress_cpu_threads=8).
Change-Id: I607a8e278e8e8f370cda512f365a97fbb81e1ae3
---
M src/kudu/integration-tests/create-table-stress-test.cc
1 file changed, 13 insertions(+), 4 deletions(-)
git pull ssh://gerrit.cloudera.org:29418/kudu refs/changes/23/11523/1
--
To view, visit http://gerrit.cloudera.org:8080/11523
To unsubscribe, visit http://gerrit.cloudera.org:8080/settings
Gerrit-Project: kudu
Gerrit-Branch: master
Gerrit-MessageType: newchange
Gerrit-Change-Id: I607a8e278e8e8f370cda512f365a97fbb81e1ae3
Gerrit-Change-Number: 11523
Gerrit-PatchSet: 1
Gerrit-Owner: Adar Dembo <[email protected]>
Gerrit-Reviewer: Alexey Serbin <[email protected]>
