HIVE-27343: Fix flaky tests: TestHiveMetaStoreTimeout#testResetTimeout #4328
Conversation
|
|
||
| // no timeout before reset | ||
| client.dropDatabase(dbName, true, true); | ||
| HMSHandler.testTimeoutValue = 250; |
There was a problem hiding this comment.
I don't get it how this config will resolve the issue. Could you please elaborate it?
There was a problem hiding this comment.
Originally the low timeout value was set at the beginning of the test. Normally it is high enough that client.dropDatabase(dbName, true, true); won't timeout, but client.createDatabase(db); will do. However, if the network/DB is very slow, even client.dropDatabase(dbName, true, true); can timeout causing the whole test to fail. Since the default value of the timeout is -1 (no timout) I moved the lowering the value after the dropdatabase call, and restored -1 after createDatabase fail, so now the test is no longer sensitive to the environment.
There was a problem hiding this comment.
The setup method of the test enables timeout. What about playing with the timeout values, just disabling timeout at the beginning of the test case and re-enable it after the drop database command?
And also, I wonder why the test testNoTimeout is not flaky. It works with the same timeout value.
There was a problem hiding this comment.
Good idea, I'll rewrite the test methods that way
kgyrtkirk
added
tests pending
tests failed
and removed
tests passed
tests pending
labels
|
LGTM. Init timeout failed at precommits so please restart the tests. Anyway, the test cases were all green so it seem we are good. |
|
Kudos, SonarCloud Quality Gate passed!
|
5 participants
What changes were proposed in this pull request?
Fix the flaky test:
TestHiveMetaStoreTimeout#testResetTimeout
Why are the changes needed?
This test was sensitive to network/DB speed, and slow CI builds were resulted in unexpected timeouts.
Does this PR introduce any user-facing change?
No
How was this patch tested?
By running TestHiveMetaStoreTimeout#testResetTimeout manually