Tim Armstrong has posted comments on this change. ( http://gerrit.cloudera.org:8080/9749 )
Change subject: IMPALA-6694: fix "buffer pool" child profile order ...................................................................... Patch Set 3: (1 comment) http://gerrit.cloudera.org:8080/#/c/9749/3/be/src/util/runtime-profile.cc File be/src/util/runtime-profile.cc: http://gerrit.cloudera.org:8080/#/c/9749/3/be/src/util/runtime-profile.cc@358 PS3, Line 358: insert_pos != children_.end() > In that case we would still get inconsistent profiles, wouldn't we? If so, In that case everything would still be correct, including the nesting structure, up to the order of children because we find the children in child_map_ and merge child profiles fine even if the order of 'children_' gets messed up. So it would really just result in a display issue. It does seem overly complicated with the two data structures to track children - we should probably just have child_map_ and a canonical order to display children in. Ah, I forgot that also with UpdateAverage() we're merging profiles from multiple executors, so it could be that they're constructed in slightly different orders. I don't know if there are any cases that we do that, but that seems even more likely than children being reordered after the first update and it's hard to guard against since most of the methods that add children are thread-safe so could be called from multiple threads. -- To view, visit http://gerrit.cloudera.org:8080/9749 To unsubscribe, visit http://gerrit.cloudera.org:8080/settings Gerrit-Project: Impala-ASF Gerrit-Branch: master Gerrit-MessageType: comment Gerrit-Change-Id: I230f0673edf20a846fdb13191b7a292d329c1bb8 Gerrit-Change-Number: 9749 Gerrit-PatchSet: 3 Gerrit-Owner: Tim Armstrong <[email protected]> Gerrit-Reviewer: Lars Volker <[email protected]> Gerrit-Reviewer: Michael Ho <[email protected]> Gerrit-Reviewer: Tim Armstrong <[email protected]> Gerrit-Comment-Date: Wed, 28 Mar 2018 05:37:04 +0000 Gerrit-HasComments: Yes
