[CI] IndexingMemoryControllerTests.testSkipRefreshIfShardIsRefreshingAlready failing occasionally

[mark-vieira]

The org.elasticsearch.indices.IndexingMemoryControllerTests. testSkipRefreshIfShardIsRefreshingAlready has started failing in CI over the past two days (beginning 9 January).

Expected: <3L>
     but: was <4L>
java.lang.AssertionError: 

Expected: <3L>
     but: was <4L>
Open stacktrace
[2020-01-10T20:07:02,409][INFO ][o.e.i.IndexingMemoryControllerTests] [testSkipRefreshIfShardIsRefreshingAlready] before test
[2020-01-10T20:07:02,661][INFO ][o.e.i.IndexingMemoryController] [testSkipRefreshIfShardIsRefreshingAlready] now throttling indexing for shard [[index][0]]: segment writing can't keep up
[2020-01-10T20:07:02,755][INFO ][o.e.i.IndexingMemoryControllerTests] [testSkipRefreshIfShardIsRefreshingAlready] after test
REPRODUCE WITH: ./gradlew ':server:test' --tests "org.elasticsearch.indices.IndexingMemoryControllerTests.testSkipRefreshIfShardIsRefreshingAlready" -Dtests.seed=B5BDCAA2A6E7D70E -Dtests.security.manager=true -Dtests.locale=vi -Dtests.timezone=Etc/GMT-7 -Dcompiler.java=13
HomeTestsorg.elasticsearch.indices.IndexingMemoryControllerTests » testSkipRefreshIfShardIsRefreshingAlready

Here are a few examples:
https://gradle-enterprise.elastic.co/s/cba7nu73aolhi/tests?search=testSkipRefreshIfShardIsRefreshingAlready
https://gradle-enterprise.elastic.co/s/ndwjucgca5frm/tests?search=testSkipRefreshIfShardIsRefreshingAlready
https://gradle-enterprise.elastic.co/s/dr4gcyuvdtmwm/tests?search=testSkipRefreshIfShardIsRefreshingAlready
https://gradle-enterprise.elastic.co/s/kkgovpbeyu34q/tests?search=testSkipRefreshIfShardIsRefreshingAlready

[mark-vieira]

Not sure what area this issue would best fall under, @dnhatn, since you last touched IndexingMemoryController maybe you might have an idea as to what is going on here, or perhaps who is best suited to look at this.

[mark-vieira]

Actually @dnhatn, looks like you just added this test yesterday so I'm just going to go ahead and assign this to you 😉

[dnhatn]

@mark-vieira Thanks for the ping. @henningandersen kindly fixed this test in #50856.