What causes "Failed to cancel query[%s] on server[%s]" in apache/druid?

Check the embedded throwable to see if the server is down (connection refused) or returned an HTTP error.

Failed to cancel query

Error message

Failed to cancel query[%s] on server[%s]

What it means

Log warning in DataServerClient when the asynchronous HTTP call to cancel a query on a data server (historical/peon) fails. The client fires cancel requests to every server hosting the query; failure to cancel on one server is logged with the cause but does not affect the cancel result returned to the caller.

Solutions

Check the embedded throwable to see if the server is down (connection refused) or returned an HTTP error.
Verify the target server is healthy and reachable (its /status endpoint).
Treat as benign if the query already completed — cancellation of finished queries commonly fails this way.
If persistent, check data server logs for the query id and inspect httpClient settings (readTimeout, numConnections) in druid.http config.

Defensive patterns

Strategy: try-catch

Validate before calling

// Pre-check server health before issuing cancel
boolean healthy = httpClient.get(serverUrl + "/status/health").isSuccess();

Try / catch

future.cancel().addListener(() -> {
  try { future.get(); }
  catch (ExecutionException e) {
    log.warn("Cancel on %s failed (may be benign if query finished): %s", location, e.getCause());
  }
});

Prevention

Treat single-server cancel failures as benign when the query already completed
Monitor data-server availability; most failures are dead/restarting historicals
Check data node logs by query id when failures are persistent

When it happens

Trigger: QueryResource.cancel POST to /druid/v2/{id} on a server returns an error or the FutureCallback's onFailure fires: server down, connection reset, query already completed on that server, or HTTP 4xx/5xx from the broker/data node.

Common situations: User cancels a query that has just finished on some servers; historical is restarting or unreachable; network partition between broker and data server; cancellation storm overloading data nodes.

Understand the failure class

Background: 'Something went wrong' / 'Request failed (500)' / 'HTTP error! status: 404' — what failed HTTP requests actually mean and how to find the real cause — this error's family across 28 libraries.

Related errors

AI-assisted analysis of apache/druid@9b90983fd2 (2026-09-07). Data as JSON: /api/errors/ae6cf1e988e1d2bd. Report an issue: GitHub.

Appendix: source

Thrown at server/src/main/java/org/apache/druid/discovery/DataServerClient.java:188

        new RequestBuilder(HttpMethod.DELETE, cancelPath).timeout(CANCELLATION_TIMEOUT),
        IgnoreHttpResponseHandler.INSTANCE
    );

    Futures.addCallback(
        cancelFuture,
        new FutureCallback<>()
        {
          @Override
          public void onSuccess(final Void result)
          {
            // Do nothing on successful cancellation.
          }

          @Override
          public void onFailure(final Throwable t)
          {
            log.noStackTrace()
               .warn(t, "Failed to cancel query[%s] on server[%s]", queryId, serviceLocation.getHostAndPort());
          }
        },
        Execs.directExecutor()
    );
  }
}

View on GitHub (pinned to 9b90983fd2)