Dinesh Bhat has posted comments on this change. Change subject: Avoid extra fsyncs of tombstoned tablets during startup ......................................................................
Patch Set 3: (1 comment) http://gerrit.cloudera.org:8080/#/c/4915/3/src/kudu/tablet/tablet_metadata.cc File src/kudu/tablet/tablet_metadata.cc: Line 173: return Status::OK(); > OK, I reorganized the code a bit in the new revision of the patch: now we c Perhaps I may be missing the entire picture here. Do we flush the tablet_data_state_ before we delete log/meta ? It appeared like with your patch earlier we were skipping tablet_data_state_ flush when a healthy but empty tablet was about to get tombstoned. If that's true, the situation I was describing was, suppose tserver receives a DeleteTablet rpc on a healthy but empty tablet (say due to a change_config triggered), then if the server crashes after it deletes the log/meta (but before we delete the superblock), then we can land up in that situation, no ? The situation being, a healthy superblock with no log/meta when tserver is coming up, not sure what state machine the tablet will go through at that point. -- To view, visit http://gerrit.cloudera.org:8080/4915 To unsubscribe, visit http://gerrit.cloudera.org:8080/settings Gerrit-MessageType: comment Gerrit-Change-Id: I60cb184b8de2a6a381371ddcf2fb938a19757eec Gerrit-PatchSet: 3 Gerrit-Project: kudu Gerrit-Branch: master Gerrit-Owner: Todd Lipcon <[email protected]> Gerrit-Reviewer: Dinesh Bhat <[email protected]> Gerrit-Reviewer: Jean-Daniel Cryans <[email protected]> Gerrit-Reviewer: Kudu Jenkins Gerrit-Reviewer: Mike Percy <[email protected]> Gerrit-Reviewer: Tidy Bot Gerrit-Reviewer: Todd Lipcon <[email protected]> Gerrit-HasComments: Yes
