Todd Lipcon has posted comments on this change. ( http://gerrit.cloudera.org:8080/9503 )
Change subject: build: adjust the number of processors for various tests ...................................................................... Patch Set 1: (5 comments) http://gerrit.cloudera.org:8080/#/c/9503/1//COMMIT_MSG Commit Message: http://gerrit.cloudera.org:8080/#/c/9503/1//COMMIT_MSG@27 PS1, Line 27: I used this as a rough guide to find tests : with inappropriate settings in CMakeLists.txt. > What's the net effect of these changes? What can I expect if I run "ctest - Done http://gerrit.cloudera.org:8080/#/c/9503/1/CMakeLists.txt File CMakeLists.txt: http://gerrit.cloudera.org:8080/#/c/9503/1/CMakeLists.txt@699 PS1, Line 699: # a heavy consumer of CPU. The PROCESSORS flag allows ctest to ensure > So the idea is that any test NOT tagged with PROCESSORS or RUN_SERIAL is a assumption is it uses 1 core. will comment http://gerrit.cloudera.org:8080/#/c/9503/1/src/kudu/util/test_main.cc File src/kudu/util/test_main.cc: http://gerrit.cloudera.org:8080/#/c/9503/1/src/kudu/util/test_main.cc@77 PS1, Line 77: const auto kInterval = MonoDelta::FromMilliseconds(500); > How did you arrive at this particular interval value? no particular reason... figured that a too-short value might over-estimate CPU consumption. If tests use a short burst of lots of CPU at startup or something it's not likely to cause other tests to fail. More concerned with if a test actually monopolizes cores for a reasonable amount of time http://gerrit.cloudera.org:8080/#/c/9503/1/src/kudu/util/test_main.cc@78 PS1, Line 78: Stopwatch sw(Stopwatch::ALL_THREADS); > What about the answers in https://stackoverflow.com/questions/12871090/how- yea, I looked at that, but then the average across the entire test lifetime is not great. I also measured things by editing run-test.sh locally to wrap the test run in 'perf stat -e task-clock -I1000 -o ${LOGFILE}.stat' and ran all the tests. That actually measures subprocesses as well, but has the disadvantage it can't be used in dist-test (since perf is not available inside our docker containers). If you think that way is better I can revert this change and instead add it as an environment variable option in run_test.sh? http://gerrit.cloudera.org:8080/#/c/9503/1/src/kudu/util/test_main.cc@104 PS1, Line 104: > Nit: revert this? Done -- To view, visit http://gerrit.cloudera.org:8080/9503 To unsubscribe, visit http://gerrit.cloudera.org:8080/settings Gerrit-Project: kudu Gerrit-Branch: master Gerrit-MessageType: comment Gerrit-Change-Id: I922031af69178b98459d0d59287442b692425afa Gerrit-Change-Number: 9503 Gerrit-PatchSet: 1 Gerrit-Owner: Todd Lipcon <t...@apache.org> Gerrit-Reviewer: Adar Dembo <a...@cloudera.com> Gerrit-Reviewer: Kudu Jenkins Gerrit-Reviewer: Todd Lipcon <t...@apache.org> Gerrit-Comment-Date: Tue, 06 Mar 2018 17:53:03 +0000 Gerrit-HasComments: Yes
