Uploaded image for project: 'MariaDB MaxScale'
  1. MariaDB MaxScale
  2. MXS-7703

exasolrouter sessions (also SmartRouter hinted to exasolrouter) do not reconnect to Exasol after an Exasol restart (error 6784)

    XMLWordPrintable

Details

    • Bug
    • Status: Closed (View Workflow)
    • Major
    • Resolution: Fixed
    • 25.10, 26.10
    • 26.10.1
    • exasolrouter
    • None
    • MXS-SPRINT-282, MXS-SPRINT-283
    • Can result in unexpected behaviour
    • Hide
      If the connection to Exasol was lost, for instance because Exasol was restarted, existing sessions of the Exasol router were never able to recover: every statement failed, even after Exasol was available again, until the client reconnected. Now the client gets the error of the failed statement, after which the session is closed, as would happen if the client were connected to a MariaDB server that went away. If the connection is lost while a large result set is being sent, the client may instead only notice that the connection is closed. A ping is now forwarded to Exasol, so it no longer succeeds when the connection has been lost. If the Exasol router is used via SmartRouter, a lost connection to a non-primary Exasol is now treated as transient, so SmartRouter can reconnect to Exasol once it is available again instead of no longer using it in that session. The same applies to the GridGain router, whose sessions previously continued silently on a new GridGain session when the connection was lost.
      Show
      If the connection to Exasol was lost, for instance because Exasol was restarted, existing sessions of the Exasol router were never able to recover: every statement failed, even after Exasol was available again, until the client reconnected. Now the client gets the error of the failed statement, after which the session is closed, as would happen if the client were connected to a MariaDB server that went away. If the connection is lost while a large result set is being sent, the client may instead only notice that the connection is closed. A ping is now forwarded to Exasol, so it no longer succeeds when the connection has been lost. If the Exasol router is used via SmartRouter, a lost connection to a non-primary Exasol is now treated as transient, so SmartRouter can reconnect to Exasol once it is available again instead of no longer using it in that session. The same applies to the GridGain router, whose sessions previously continued silently on a new GridGain session when the connection was lost.

    Description

      Summary

      After the Exasol backend is restarted, client sessions that were already open on the exasolrouter (directly, or through SmartRouter with a hint that routes to exasolrouter) never recover. Every query on such a session fails with error 6784 even though Exasol is accepting connections again. A new client connection works immediately.

      Environment

      • MaxScale 26.10.0-rc1 (image us-central1-docker.pkg.dev/mariadb-es-docker-registry/docker-proxy-dev/maxscale:26.10.0-rc1)
      • MariaDB 11.8, Exasol docker-db 2026.1.0
      • exasolrouter listener on :3311, SmartRouter on :3310

      Steps to reproduce

      1. Open a long-lived client session on the exasolrouter listener (:3311), and a second one on SmartRouter (:3310) with the hint /* maxscale route to server exasolrouter */.
      2. On each, run select count from flights.flights where id > N every 5 seconds.
      3. Restart the Exasol container (it is back and accepting SQL after about 33 seconds).

      Observed

      • While Exasol is down: ERROR 49215 (08S01): [Exasol][Exasol Driver]No server listening.
      • After Exasol is back, every later query on the same session fails: ERROR 6784 (HY000): [Exasol][Exasol Driver]Database server not connected.
        • :3311 session: still failing about 2m40s after Exasol returned (34 consecutive failures).
        • :3310 session with the exasolrouter hint: still failing about 2m17s after Exasol returned (32 consecutive failures).
      • A new connection to the same listener immediately returns correct results, so reconnecting the client is the only workaround seen.
      • An unhinted SmartRouter session kept working through the restart, but only because MariaDB answered every query (Exasol was never involved). Exasol-routed reads through SmartRouter were not tested.

      Not tested

      • Waits longer than about 3 minutes after Exasol is back.
      • Any MaxScale timeout or reconnect settings.

      Expected

      exasolrouter re-opens its ODBC connection to Exasol after a failure, or closes the client session so that clients reconnect.

      Origin

      Reported from a MariaDB Cloud deployment: after deleting the Exasol pod and letting it rebuild, existing connections return errors and have to be recreated before data can be queried again.

      Created via the jira-create skill.

      Attachments

        Issue Links

          Activity

            People

              johan.wikman Johan Wikman
              allen.herrera Allen Herrera
              Votes:
              0 Vote for this issue
              Watchers:
              1 Start watching this issue

              Dates

                Created:
                Updated:
                Resolved:

                Time Tracking

                  Estimated:
                  Original Estimate - 1d
                  1d
                  Remaining:
                  Remaining Estimate - 0d
                  0d
                  Logged:
                  Time Spent - 7h Time Not Required
                  7h

                  Git Integration

                    Error rendering 'com.xiplink.jira.git.jira_git_plugin:git-issue-webpanel'. Please contact your Jira administrators.