Quanlong Huang has posted comments on this change. ( 
http://gerrit.cloudera.org:8080/21573 )

Change subject: IMPALA-13208: Add cluster id to the membership and 
request-queue topic names
......................................................................


Patch Set 3:

(1 comment)

http://gerrit.cloudera.org:8080/#/c/21573/3//COMMIT_MSG
Commit Message:

http://gerrit.cloudera.org:8080/#/c/21573/3//COMMIT_MSG@9
PS3, Line 9: To share catalogd and statestore across Impala clusters, this adds 
the
> How will this all interact with admission control?
The membership and request-queue topics are indeed used by admission control to 
schedule queries. The difference to executor groups is that planner is aware of 
difference executor groups. Coordinator plans and schedules queries to 
different executor groups. The coordinator is shared between executor groups. 
They are still an Impala cluster.

This patch makes different Impala clusters be able to share the same catalogd 
and statestore service. But each cluster still has their own coordinators.

In the future when we ensure the backward compatibility of catalog and 
statestore services, clusters of different versions can also share the same 
catalogd and statestored. Admins will be able to upgrade individual clusters.

This also saves the cost of deploying catalogd for each cluster, and makes the 
metadata sync between clusters easier since they are using the same catalogd.



--
To view, visit http://gerrit.cloudera.org:8080/21573
To unsubscribe, visit http://gerrit.cloudera.org:8080/settings

Gerrit-Project: Impala-ASF
Gerrit-Branch: master
Gerrit-MessageType: comment
Gerrit-Change-Id: I2ff41539f568ef03c0ee2284762b4116b313d90f
Gerrit-Change-Number: 21573
Gerrit-PatchSet: 3
Gerrit-Owner: Quanlong Huang <[email protected]>
Gerrit-Reviewer: Impala Public Jenkins <[email protected]>
Gerrit-Reviewer: Michael Smith <[email protected]>
Gerrit-Reviewer: Quanlong Huang <[email protected]>
Gerrit-Comment-Date: Mon, 15 Jul 2024 06:12:50 +0000
Gerrit-HasComments: Yes

Reply via email to