Adar Dembo has posted comments on this change. Change subject: process_memory: incrementally track total process consumption ......................................................................
Patch Set 1: (3 comments) http://gerrit.cloudera.org:8080/#/c/6793/1//COMMIT_MSG Commit Message: PS1, Line 9: This switches CurrentConsumption() from calling into TCMalloc to instead : using its own tracking, based on installing new/delete hooks into : tcmalloc. The hooks incrementally update a global LongAdder for : consumption, and we just query that object when we need to know the : current value. Is there any deviation between our tracked value and the TCMalloc reported value? Would be interesting to compare the two on a long-running (or heavily used) server, and perhaps enforce via test in some way. PS1, Line 28: On the side, this changes process_memory to use GoogleOnce instead of : std::once. This sped up the microbenchmark significantly (almost 2x), : apparently due to better inlining. This is somewhat concerning. I presume your benchmark was on el6? Wondering whether it's still true with a more modern libstdc++; can you try on Ubuntu 16.04? http://gerrit.cloudera.org:8080/#/c/6793/1/src/kudu/util/CMakeLists.txt File src/kudu/util/CMakeLists.txt: Line 356: ADD_KUDU_TEST(process_memory-test RUN_SERIAL true) Nit: out of order. -- To view, visit http://gerrit.cloudera.org:8080/6793 To unsubscribe, visit http://gerrit.cloudera.org:8080/settings Gerrit-MessageType: comment Gerrit-Change-Id: I3293110b7eda7f083bf400e165be3d782c441c49 Gerrit-PatchSet: 1 Gerrit-Project: kudu Gerrit-Branch: master Gerrit-Owner: Todd Lipcon <[email protected]> Gerrit-Reviewer: Adar Dembo <[email protected]> Gerrit-Reviewer: Jean-Daniel Cryans <[email protected]> Gerrit-Reviewer: Kudu Jenkins Gerrit-HasComments: Yes
