Increase query timeout in testDeleteRowsConcurrently #20328
Conversation
findepi
added
test
no-release-notes
| @@ -156,7 +157,9 @@ public void testDeleteRowsConcurrently() | |||
| .collect(toImmutableList()); | |||
|
|
|||
| Stream<Optional<String>> expectedRows = Streams.mapWithIndex(futures.stream(), (future, index) -> { | |||
There was a problem hiding this comment.
Let's rather reduce the concurrency
nThreads = 3
TestTable table = new TestTable(
getQueryRunner()::execute,
"test_concurrent_delete",
"(col0 INTEGER, col1 INTEGER, col2 INTEGER)")) {
The way this test is conceived, it can get into a metastore replacement operation failure spiral and there's nothing that Trino (at the moment) can do about it.
There was a problem hiding this comment.
what is "metastore replacement operation failure spiral"?
There was a problem hiding this comment.
Potentially trying to update repeatedly the metastore entry at the same time repeatedly
There was a problem hiding this comment.
how long can it reasonably take for 4 queries? and how long in extreme case? what would be safe test timeout?
how does the number look like with 3 queries?
There was a problem hiding this comment.
#13995 states
Error: io.trino.plugin.iceberg.TestIcebergGcsConnectorSmokeTest.testDeleteRowsConcurrently Time elapsed: 120.014 s <<< FAILURE!
so - in some peculiar cases it seems to take more than 2 minutes. i don't see any value in investing over 2 minutes to test this functionality.
There was a problem hiding this comment.
in that issue there are two failure modes
overall test timeout after 120s. last known occurrence April 2023 #13995 (comment)
timeout when waiuting for individual query to complete (increaeased here) "NoSuchElementException: No value present" -- last known occurence Aug 2023 #13995 (comment) and today
this is this second problem that i am trying to change
i don't feel strongly though, if you have a better idea how to help this test fail less frequently
There was a problem hiding this comment.
ok. let's fix the second problem.
if the timeout will happen again we can reduce the concurrency.
findepi
force-pushed
the
findepi/increase-query-timeout-in-testdeleterowsconcurrently-f9506a
branch
from
60664a2 to
dbe9067
Compare
|
CI #20326 |
findepi
deleted the
findepi/increase-query-timeout-in-testdeleterowsconcurrently-f9506a
branch
Maybe fixes #13995