Commit Graph

2 Commits

Author SHA1 Message Date
Azat Khuzhin
205ebc7cff Tune idle_connection_timeout to avoid reconnects
There may be 1 hour between two different distributed queries in tests
(see [1] and [2]), if this will happen then the connection may be closed
already due to idle_connection_timeout.

  [1]: https://clickhouse-test-reports.s3.yandex.net/21432/572d5862961ffda88125679e3a5ea684c316b223/functional_stateless_tests_(debug).html#fail1
  [2]: https://clickhouse-test-reports.s3.yandex.net/21432/572d5862961ffda88125679e3a5ea684c316b223/functional_stateless_tests_(antlr_debug).html#fail1

Let's tune it to avoid such issues and override other settings in tests.

Another option is to execute SYSTEM RELOAD CONFIG before each test...
2021-03-04 22:48:09 +03:00
Azat Khuzhin
7794ac66d4 Fix flaky tests with distributed queries
There was few attempts to fix this [1] and [2], but it still pops up
sometimes, for example here [3].

  [1]: 009f57fc2f / #11211
  [2]: cdb6bed8b9 / #14198
  [3]: https://clickhouse-test-reports.s3.yandex.net/21318/38be9ff43ac4c46ce6e803fc125d910bde1d4c71/functional_stateless_tests_(release,_databasereplicated).html#fail1

Let's use more generic approach, and do not hide any errors (and see
does increasing timeout will be enough).
2021-03-04 22:48:04 +03:00