Adar Dembo has posted comments on this change. Change subject: Add a simple benchmark to create 1M blocks and reopen LBM ......................................................................
Patch Set 2: Code-Review+2 (1 comment) Based on my understanding of mkfs block size choices, I think your assertion about sync_file_range and stable page writes is overly strong. But the code is fine, so feel free to ignore and merge as is. http://gerrit.cloudera.org:8080/#/c/8006/2/src/kudu/fs/log_block_manager-test.cc File src/kudu/fs/log_block_manager-test.cc: PS2, Line 797: In particular, : // since we write such small blocks in this test, each block will likely : // begin on the same 4KB page as the prior one we wrote, and due to the : // "stable page writes" feature, each block will thus end up waiting : // on the writeback of the prior one. That's only true when the filesystem block size is less than 4K though, right? I don't know why many of our test machines wound up with 2K ext4 filesystems; 4K is generally the default on ext4 (provided the fs isn't of a trivially small size). -- To view, visit http://gerrit.cloudera.org:8080/8006 To unsubscribe, visit http://gerrit.cloudera.org:8080/settings Gerrit-MessageType: comment Gerrit-Change-Id: I5cc3547db7d8389c4e89ff9d4d3043b5f2fbe878 Gerrit-PatchSet: 2 Gerrit-Project: kudu Gerrit-Branch: master Gerrit-Owner: Todd Lipcon <[email protected]> Gerrit-Reviewer: Adar Dembo <[email protected]> Gerrit-Reviewer: Dan Burkert <[email protected]> Gerrit-Reviewer: David Ribeiro Alves <[email protected]> Gerrit-Reviewer: Hao Hao <[email protected]> Gerrit-Reviewer: Kudu Jenkins Gerrit-Reviewer: Todd Lipcon <[email protected]> Gerrit-HasComments: Yes
