Adar Dembo has posted comments on this change. ( http://gerrit.cloudera.org:8080/13975 )
Change subject: KUDU-2911: Consider the available space when selecting data dirs for tablets ...................................................................... Patch Set 4: (1 comment) http://gerrit.cloudera.org:8080/#/c/13975/3/src/kudu/fs/data_dirs.cc File src/kudu/fs/data_dirs.cc: http://gerrit.cloudera.org:8080/#/c/13975/3/src/kudu/fs/data_dirs.cc@988 PS3, Line 988: int uuid_idx = (*group_uuid_indices)[i]; > Thanks a lot. I have restored it to the original. This made me very trouble It looks like PS4 works like this: 1. When creating a new tablet, take available disk space into account when choosing data dirs. 2. When creating a new block, continue to randomly select amongst the tablet's data dirs for the new block's home. IIUC #2 was previously addressed by PS3, but that was reverted for PS4. And that's what Andrew had commented on (i.e. all the blocks stacking on one disk because it had more available space). I like Andrew's idea of extending power-of-2-choice to this problem. Would it work like this? 1. First choose two data dirs at random from the tablet's data dir group. 2. Then choose whichever of the two data dirs has more free space. Would it be possible to profile the performance difference, perhaps using YCSB? -- To view, visit http://gerrit.cloudera.org:8080/13975 To unsubscribe, visit http://gerrit.cloudera.org:8080/settings Gerrit-Project: kudu Gerrit-Branch: master Gerrit-MessageType: comment Gerrit-Change-Id: I194c4965ee64aed728e3b84e684c04d445cbe529 Gerrit-Change-Number: 13975 Gerrit-PatchSet: 4 Gerrit-Owner: ZhangYao <triplesheep0...@gmail.com> Gerrit-Reviewer: Adar Dembo <a...@cloudera.com> Gerrit-Reviewer: Andrew Wong <aw...@cloudera.com> Gerrit-Reviewer: Kudu Jenkins (120) Gerrit-Reviewer: Yingchun Lai <405403...@qq.com> Gerrit-Reviewer: ZhangYao <triplesheep0...@gmail.com> Gerrit-Comment-Date: Fri, 02 Aug 2019 04:14:41 +0000 Gerrit-HasComments: Yes