Jason Fehr has posted comments on this change. ( 
http://gerrit.cloudera.org:8080/22249 )

Change subject: IMPALA-13201: System Table Queries Execute When Admission 
Queues are Full
......................................................................


Patch Set 20:

(7 comments)

http://gerrit.cloudera.org:8080/#/c/22249/19/be/src/scheduling/cluster-membership-mgr-test.cc
File be/src/scheduling/cluster-membership-mgr-test.cc:

http://gerrit.cloudera.org:8080/#/c/22249/19/be/src/scheduling/cluster-membership-mgr-test.cc@290
PS19, Line 290:       << "Actual hostnames:" << actual_hostnames.str();
> Space after "Actual hostnames:" isn't needed because actual_hostnames will
Done


http://gerrit.cloudera.org:8080/#/c/22249/19/be/src/scheduling/cluster-membership-mgr-test.cc@302
PS19, Line 302:         << expected << "' in actual coordinators:" << 
actual_hostnames.str();
> Same here, could be "' in actual coordinators:".
Done


http://gerrit.cloudera.org:8080/#/c/22249/15/be/src/util/webserver.cc
File be/src/util/webserver.cc:

http://gerrit.cloudera.org:8080/#/c/22249/15/be/src/util/webserver.cc@717
PS15, Line 717:             LOG(INFO) << "Invalid JWT token provided";
> Eh, guess it's a good idea to clean up.
Done


http://gerrit.cloudera.org:8080/#/c/22249/19/common/thrift/ImpalaInternalService.thrift
File common/thrift/ImpalaInternalService.thrift:

http://gerrit.cloudera.org:8080/#/c/22249/19/common/thrift/ImpalaInternalService.thrift@236
PS19, Line 236:   // Specifies the state of the onlyCoordinators configuration 
element for a request pool.
> Should have a comment.
Good catch.  Done


http://gerrit.cloudera.org:8080/#/c/22249/19/docs/topics/impala_admission_config.xml
File docs/topics/impala_admission_config.xml:

http://gerrit.cloudera.org:8080/#/c/22249/19/docs/topics/impala_admission_config.xml@187
PS19, Line 187:           To use an <codeph>&lt;onlyCoordinators&gt;</codeph> 
request pool, set the
> typo: "simple" should be "simply", but I'd rather remove it.
Done


http://gerrit.cloudera.org:8080/#/c/22249/19/docs/topics/impala_admission_config.xml@196
PS19, Line 196:           Caution: care must be taken when naming the 
<codeph>&lt;onlyCoordinators&gt;</codeph>
> Why does this happen?
It's a combination of planning and scheduling.  Impala allows for specifying 
the "expected_executor_group_sets" flag which defines groups of isolated 
executor groups that the planner and scheduler then select from when a query is 
to be executed.

For example, say there are executor group set "small" and "large".  Executor 
groups "small-001", "small-002", and "small-003" will all be considered by the 
scheduler when the planner determines that executor group set "small" can run 
they query.  Similarly, if the planner determines that "small" is too small, a 
a "large" executor group needs to run the query, then the scheduler will 
consider executor groups "large-001", "large-002", and "large-003".

If an only coordinator's pool is named "small-onlycoords", it will be 
considered when the planner selects "small" as the executor group set.

I looked into erroring out if an only coordinators request pool overlaps with 
an executor group set, but it's unfortunately not easy to do that because the 
executor group sets are maintained in the backend while the request pools are 
maintained in the frontend.


http://gerrit.cloudera.org:8080/#/c/22249/16/tests/custom_cluster/test_admission_controller.py
File tests/custom_cluster/test_admission_controller.py:

http://gerrit.cloudera.org:8080/#/c/22249/16/tests/custom_cluster/test_admission_controller.py@1949
PS16, Line 1949:     self._start_impala_cluster(
               :         options=[
               :             
"--impalad_args=s{}".format(impalad_admission_ctrl_config_args(
               :                 
fs_allocation_file="fair-scheduler-onlycoords.xml",
               :                 llama_site_file="llama-site-onlycoords.xml"))],
               :         add_impalads=True,
               :         cluster_size=6,
               :         num_coordinators=1,
               :         use_exclusive_coordinators=True,
               :         wait_for_backends=False,
               :         workload_mgmt=True)
> What you're running up against here is that add_impalads was hacked in to s
https://issues.apache.org/jira/browse/IMPALA-13755



--
To view, visit http://gerrit.cloudera.org:8080/22249
To unsubscribe, visit http://gerrit.cloudera.org:8080/settings

Gerrit-Project: Impala-ASF
Gerrit-Branch: master
Gerrit-MessageType: comment
Gerrit-Change-Id: I5e0e64db92bdbf80f8b5bd85d001ffe4c8c9ffda
Gerrit-Change-Number: 22249
Gerrit-PatchSet: 20
Gerrit-Owner: Jason Fehr <[email protected]>
Gerrit-Reviewer: Abhishek Rawat <[email protected]>
Gerrit-Reviewer: Andrew Sherman <[email protected]>
Gerrit-Reviewer: Impala Public Jenkins <[email protected]>
Gerrit-Reviewer: Jason Fehr <[email protected]>
Gerrit-Reviewer: Joe McDonnell <[email protected]>
Gerrit-Reviewer: Michael Smith <[email protected]>
Gerrit-Reviewer: Riza Suminto <[email protected]>
Gerrit-Comment-Date: Wed, 12 Feb 2025 00:20:25 +0000
Gerrit-HasComments: Yes

Reply via email to