[jira] [Created] (IMPALA-13366) Query stuck in the queue during graceful shutdown

["Yida Wu (Jira)" <ji...@apache.org>]

Yida Wu created IMPALA-13366:
--------------------------------

             Summary: Query stuck in the queue during graceful shutdown
                 Key: IMPALA-13366
                 URL: https://issues.apache.org/jira/browse/IMPALA-13366
             Project: IMPALA
          Issue Type: Bug
          Components: Backend
    Affects Versions: Impala 4.4.1
            Reporter: Yida Wu


There is an issue during graceful shutdown when a query is in the queue. The queued query is not executed, and can cause the shutdown process to be blocked until it hits queue_wait_timeout_ms. To reproduce this behavior:
1. Set the flag default_pool_max_requests=1, for example,
bq. $IMPALA_HOME/bin/start-impala-cluster.py \-\-impalad_args="--default_pool_max_requests=1" -s 1
2. Use client A to submit a query: select sleep(20000).
3. While the query from step 2 is still running, use a different client B to submit another query: select sleep(1).
4. Run "kill -SIGRTMIN $(pgrep impalad)"
In this case, only one running query is allowed, and the query from step 3 will be queued. After the query from step 2 completes, the query from step 3 will remain queued and won't execute until it times out based on queue_wait_timeout_ms.



--
This message was sent by Atlassian Jira
(v8.20.10#820010)