Details
-
Bug
-
Status: Resolved
-
Major
-
Resolution: Cannot Reproduce
-
None
-
None
-
ghx-label-6
Description
At times, observed that queries has been queued when pool is idle - no other queries are running in the pool. There are couple of incidents. Observation for one of the incidents is as follows:
Had come across a situation where in queue started building up on specific pool. On analysis, come to know that queuing happens only when query is hitting 1 specific node. Then executed a simple query directly (using impala shell) on that node has been queued, whereas, same query executed in other nodes are able to admit immediately and completes its execution even though both these runs are using the same pool. I think that problematic node had this problem and then it is clearing off automatically, but am not sure. But in this case, restarted daemon. I suspect somehow admission control accountability info (How much has used? How much left? etc) is not available in this node for some reasons. Please share your view on this.