Bikramjeet Vig has posted comments on this change. ( http://gerrit.cloudera.org:8080/17209 )
Change subject: IMPALA-10594: Handle failed coordinators in admissiond ...................................................................... Patch Set 2: (1 comment) http://gerrit.cloudera.org:8080/#/c/17209/2//COMMIT_MSG Commit Message: http://gerrit.cloudera.org:8080/#/c/17209/2//COMMIT_MSG@13 PS2, Line 13: its also possible that the coordinator probably never went down but due to network issues the statestore marks it as down. In this case there can be fragments running on the other coordinators/executors and removing those queries can result in overadmission on the ones that are running those fragments. This should ideally be prevented by the host level memory accounting in admission controller, but I need to re-check if it works here. Also once the coord comes back up on the statestore, what will happen to the already running queries?? do they get accounted for again and potentially cause overadmission? -- To view, visit http://gerrit.cloudera.org:8080/17209 To unsubscribe, visit http://gerrit.cloudera.org:8080/settings Gerrit-Project: Impala-ASF Gerrit-Branch: master Gerrit-MessageType: comment Gerrit-Change-Id: I883f323bb765680ef24b3c3f51fb209dea15f0b0 Gerrit-Change-Number: 17209 Gerrit-PatchSet: 2 Gerrit-Owner: Thomas Tauber-Marshall <[email protected]> Gerrit-Reviewer: Bikramjeet Vig <[email protected]> Gerrit-Reviewer: Impala Public Jenkins <[email protected]> Gerrit-Reviewer: Wenzhe Zhou <[email protected]> Gerrit-Comment-Date: Fri, 19 Mar 2021 21:00:18 +0000 Gerrit-HasComments: Yes
