Bikramjeet Vig created IMPALA-9077:
--------------------------------------
Summary: Fix scalable admission config for clusters with dedicated
coordinators
Key: IMPALA-9077
URL: https://issues.apache.org/jira/browse/IMPALA-9077
Project: IMPALA
Issue Type: Bug
Affects Versions: Impala 3.3.0
Reporter: Bikramjeet Vig
Assignee: Bikramjeet Vig
The scalable configs are:
Max Memory Multiple
Max Running Queries Multiple
Max Queued Queries Multiple
Currently the scalable configs scale based on number of total executors. This
can work for max running or queued queries multiple, but for "Max Memory
Multiple" this will cause the following problem:
Consider the following cluster composed of 2 dedicated coordinators and 2
executors. For a regular query that will run on 2 executors and 1 coordinator,
if the memory to admit per backend is 100MB, the cluster memory to be admitted
will be 300MB. Now if the max_memory_multiple is 120MB this query will not be
admitted because the scalable max memory will be 120*2 = 240.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
