sql/logictest: TestLogic/local/order_by timed out under race

[cockroach-teamcity]

(sql/logictest).TestLogic/local/order_by failed on release-20.2@918930e2a9e722943c50e77519e6dea93aed20e7:

Slow failing tests:
TestLogic/local/order_by - 0.32s

Slow passing tests:
<none>

More

make stressrace TESTS=TestLogic/local/order_by PKG=./pkg/sql/logictest TESTTIMEOUT=5m STRESSFLAGS='-timeout 5m' 2>&1

See this test on roachdash
_{powered by pkg/cmd/internal/issues}

[asubiotto]

cc @rytaft any idea why this didn't run with your new 2h timeout? My hunch is that teamcity uses the script on the branch and one of #54615 / #54691 need to be backported.

[rytaft]

Yea, I think I need to backport the changes. But unfortunately I still haven't found a configuration that will succeed on master. It's timing out after 4 hours on TestLogic/local/aggregate/other. I'm having a hard time understanding whether that's expected with the race build, or if it's actually getting blocked on something.

It's failing consistently in the same spot after 154 statements/queries, which if I'm counting correctly in the file seems to be this:

statement ok
INSERT INTO mnop (m, n) SELECT i, (1e9 + i/2e4)::float FROM
  generate_series(1, 2e4) AS i(i)

If that's the problematic statement, it's not that surprising since it's performing a lot of numeric computation while inserting 20,000 rows. I know that logic tests have been excluded from race builds in the past, so maybe this is why? Maybe I can just exclude the aggregate file?

Anyway, you can definitely feel free to close this issue -- sorry for the spam.

[asubiotto]

I took a look at that test in #54685 and it seems like a rangefeed is getting stuck. Going to close this issue.

[rytaft]

Great - thanks. In that case, I'll backport those PRs and maybe the rangefeed issue won't be a problem on the older branches.