Philip Zeyliger has posted comments on this change. ( http://gerrit.cloudera.org:8080/9395 )
Change subject: IMPALA-6341, IMPALA-5917: Reduce mem-limit for start-impala-cluster. ...................................................................... Patch Set 1: I see this happening in multiple environments, but the environment I see this most uses a little bit more memory for HiveServer2 than other places. (I have another change that tries to get rid of Yarn which would free up some space.) The math here is fundamentally wrong. If we wanted to do it correctly, we'd say: whole_machine = X GB minicluster = Y GB what's left = X - Y random_spare_space = 2 GB per-impalad: (X - Y - 2) / 3 And, finally, we'd assert that per-impalad is big enough. I've added enough monitoring in https://gerrit.cloudera.org/#/c/9291/ to eventually have enough data to make this more explicit, but given that we've muddled along similarly for a while, I don't think it's worth embarking on a bigger effort like above. -- To view, visit http://gerrit.cloudera.org:8080/9395 To unsubscribe, visit http://gerrit.cloudera.org:8080/settings Gerrit-Project: Impala-ASF Gerrit-Branch: master Gerrit-MessageType: comment Gerrit-Change-Id: Iabca7a95560bd27c2de2b0a147ee9a3c45199db7 Gerrit-Change-Number: 9395 Gerrit-PatchSet: 1 Gerrit-Owner: Philip Zeyliger <phi...@cloudera.com> Gerrit-Reviewer: Alex Behm <alex.b...@cloudera.com> Gerrit-Reviewer: Philip Zeyliger <phi...@cloudera.com> Gerrit-Comment-Date: Thu, 22 Feb 2018 17:31:35 +0000 Gerrit-HasComments: No
