JAVA-2841: Raise timeouts during connection initialization

[tomekl007]

I've tested the new timeout by connecting to Astra cluster that is in us-west from central Europe using this code:

public class AstraTest {
  public static void main(String[] args) {

    for (int i = 0; i < 100; i++) {
      Path bundle = Paths.get("/Users/tomaszlelek/Downloads/secure-connect-cass-dc.zip");

      try (CqlSession session =
          CqlSession.builder()
              .withCloudSecureConnectBundle(bundle)
              .withAuthCredentials("datastax", "datastax")
              .build()) {
        ResultSet set = session.execute("select * from system.local");
        System.out.println(set.all());
      }
    }
  }
}

with the previous timeout of 500 milliseconds, it was constantly failing to connect.
For the new timeout of 5 seconds, it has successfully connected 100 times.
I don't see a clear reason to increase connect-timeout timeout as well - @adutra can you elaborate more on that suggestion?

[adutra]

@olim7t knows better than me but afaik connect-timeout is for the duration of the whole connection opening process, whereas init-query-timeout applies to each individual request made during connection initialization. IOW, connect-timeout is necessarily much larger than init-query-timeout since it encompasses many init queries. That's why I figured that setting both timeouts to the same value doesn't make sense.

[olim7t]

Actually connect-timeout only applies to opening the socket, it does not span protocol negotiation. So I think the current value is fine.

[olim7t]

Don't forget the changelog entry.

nit: could you follow the project's branch naming convention for future PRs? java2841

[olim7t]

reference.conf uses HOCON substitutions to define a few other options in terms of this one. You need to update them too: CONNECTION_SET_KEYSPACE_TIMEOUT, HEARTBEAT_TIMEOUT, CONTROL_CONNECTION_TIMEOUT and REPREPARE_TIMEOUT.

In fact we should probably extract a variable:

    Duration initQueryTimeout = Duration.ofSeconds(5);
    map.put(TypedDriverOption.CONNECTION_INIT_QUERY_TIMEOUT, initQueryTimeout);
    map.put(TypedDriverOption.CONNECTION_SET_KEYSPACE_TIMEOUT, initQueryTimeout);
    ...

Can you also do a pass to handle all other substitutions the same way, and commit that separately?

MapBasedDriverConfigLoaderTest.should_fill_default_profile_like_reference_file will let you know if everything is in sync.

[tomekl007]

done, what do you mean in a separate commit? Do you want to have two commits in the 4.x after this branch will be merged?

[adutra]

Yes, because the refactoring to handle other substitutions by extracting a local variable is unrelated to this PR.

[adutra]

looks good, but as Olivier suggested, please do a pass to extract local variables for similar cases of hocon references in OptionsMap.

[tomekl007]

looks good, but as Olivier suggested, please do a pass to extract local variables for similar cases of hocon references in OptionsMap.

It is already done in:
acc08c0

[adutra]

It is already done in: acc08c0

No, I believe he was asking to apply the same refactoring to other cases of hocon references found in reference.conf.

[tomekl007]

It is already done in: acc08c0

No, I believe he was asking to apply the same refactoring to other cases of hocon references found in reference.conf.

You are right, done in 7ad50aa