Bikramjeet Vig has posted comments on this change. ( http://gerrit.cloudera.org:8080/15715 )
Change subject: IMPALA-9643: fix runtime filter race for mt_dop ...................................................................... Patch Set 5: (2 comments) http://gerrit.cloudera.org:8080/#/c/15715/4/be/src/runtime/runtime-filter-bank.cc File be/src/runtime/runtime-filter-bank.cc: http://gerrit.cloudera.org:8080/#/c/15715/4/be/src/runtime/runtime-filter-bank.cc@118 PS4, Line 118: ret = obj_pool_.Add(new RuntimeFilter(filter_desc, filter_desc.filter_size_bytes)); any reason we lazily create these filters when instances register? We actually already know during RuntimeFilterBank creation which filters will be generated and which fragments will be executed. can we infer from that info about which filters need to allocate consumed_filter and produced_filter and avoid this problem since UpdateFilterFromLocal will always have an available filter to set. http://gerrit.cloudera.org:8080/#/c/15715/4/be/src/runtime/runtime-filter-bank.cc@128 PS4, Line 128: filter_desc.is_broadcast_join is it possible to also hit the same issue with non-broadcast joins. It will be less likely since all producers need to publish but we can still hit that case though right? -- To view, visit http://gerrit.cloudera.org:8080/15715 To unsubscribe, visit http://gerrit.cloudera.org:8080/settings Gerrit-Project: Impala-ASF Gerrit-Branch: master Gerrit-MessageType: comment Gerrit-Change-Id: I194c0d2515b6a0e5474e1c0c8647f0e54dc94397 Gerrit-Change-Number: 15715 Gerrit-PatchSet: 5 Gerrit-Owner: Tim Armstrong <[email protected]> Gerrit-Reviewer: Bikramjeet Vig <[email protected]> Gerrit-Reviewer: Impala Public Jenkins <[email protected]> Gerrit-Reviewer: Thomas Tauber-Marshall <[email protected]> Gerrit-Reviewer: Tim Armstrong <[email protected]> Gerrit-Comment-Date: Mon, 13 Apr 2020 20:33:00 +0000 Gerrit-HasComments: Yes
