HIVE-29513: Increase HikariCP connection timeout default from 30s to 60s

[rtrivedi12]

What changes were proposed in this pull request?

Set HikariCP connection timeout to an explicit default of 60 seconds (60,000 ms) in HikariCPDataSourceProvider, making it configurable via hikaricp.connectionTimeout property.

Why are the changes needed?

MetaStore experiences connection timeout errors when database response is slow or during peak load, as HikariCP's implicit 30-second default is insufficient.

Does this PR introduce any user-facing change?

Yes. Connection timeout default changes from HikariCP's implicit 30 seconds to Hive's explicit 60 seconds. Users can override this by setting hikaricp.connectionTimeout in hive-site.xml.

How was this patch tested?

Added/updated unit tests in TestDataSourceProviderFactory:

[ayushtkn]

I have a minor refactor suggestion, otherwise change LGTM

[ayushtkn]

Can we instead handle it in the getPrefixedProperties itself?

  static Properties getPrefixedProperties(Configuration hdpConfig, String prefix, Map<String, Long> timeDurationConfigs) {
    Properties props = new Properties();

    Iterables.filter(hdpConfig, e -> e.getKey() != null && e.getKey().startsWith(prefix)).forEach(e -> {
      String fullKey = e.getKey();
      String keyName = fullKey.substring(prefix.length());
      Long defaultVal = timeDurationConfigs.get(keyName);
      if (defaultVal != null) {
        long timeMs = hdpConfig.getTimeDuration(fullKey, defaultVal, TimeUnit.MILLISECONDS);
        props.setProperty(fullKey, String.valueOf(timeMs));
      } else {
        props.setProperty(fullKey, e.getValue());
      }
    });

    return props;
  }

and define a MAP like

  private static final Map<String, Long> TIME_DURATION_CONFIGS = Map.of("connectionTimeout",
      DEFAULT_CONNECTION_TIMEOUT_MS);

[rtrivedi12]

Sure, thanks for suggestion!

[okumin]

Does it often happen that the backend RDBMS is unavailable for 30-60 seconds?

[rtrivedi12]

Does it often happen that the backend RDBMS is unavailable for 30-60 seconds?

Yes, we've mainly observed connection drops during peak traffic or heavy transactional table usage. However, increasing the connection timeout to 60 seconds generally allows the system to recover sucessfully.

[sonarqubecloud]

Quality Gate passed

Issues
0 New issues
0 Accepted issues

Measures
0 Security Hotspots
0.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud

[okumin]

In my honest opinion, the configuration should be tuned by users, or we should drastically increase the default if it is evident that typical connection pool settings don't fit HMS's workload.
If it takes 40 seconds today in someone's environment today, it might take 61 seconds on another day or in another environment. It might be too ad-hoc to update the default timeout every time someone hits the issue.

[github-actions]

This pull request has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs.
Feel free to reach out on the dev@hive.apache.org list if the patch is in need of reviews.