Adar Dembo has posted comments on this change. ( http://gerrit.cloudera.org:8080/13820 )
Change subject: [docs] update the upgrade documentation ...................................................................... Patch Set 2: (11 comments) http://gerrit.cloudera.org:8080/#/c/13820/2/docs/installation.adoc File docs/installation.adoc: http://gerrit.cloudera.org:8080/#/c/13820/2/docs/installation.adoc@635 PS2, Line 635: for the scenario of compiling from source code. "relevant when building from source code". http://gerrit.cloudera.org:8080/#/c/13820/2/docs/installation.adoc@640 PS2, Line 640: kudu-tserver "tablet server" http://gerrit.cloudera.org:8080/#/c/13820/2/docs/installation.adoc@640 PS2, Line 640: 2hours "2 hours" http://gerrit.cloudera.org:8080/#/c/13820/2/docs/installation.adoc@645 PS2, Line 645: --force Shouldn't need this; the flag is tagged as 'runtime'. http://gerrit.cloudera.org:8080/#/c/13820/2/docs/installation.adoc@647 PS2, Line 647: and the gflag above should be reset after every reboot. Since step 4 restores the gflag value, do we really need this step? http://gerrit.cloudera.org:8080/#/c/13820/2/docs/installation.adoc@647 PS2, Line 647: kudu-tserver "the tablet servers" http://gerrit.cloudera.org:8080/#/c/13820/2/docs/installation.adoc@649 PS2, Line 649: [NOTE] Shouldn't this be just after "Set the unavailable time..." so users understand why it's important to change the gflag value before the rolling restart? http://gerrit.cloudera.org:8080/#/c/13820/2/docs/installation.adoc@651 PS2, Line 651: If the gflag is not reset, kudu-tserver which is restarting would be possibly evicted from the cluster. "If the unavailable time is not extended, restarted tablet servers could be evicted from the cluster." http://gerrit.cloudera.org:8080/#/c/13820/2/docs/installation.adoc@654 PS2, Line 654: Rolling restart We should define what "rolling restart" means too. http://gerrit.cloudera.org:8080/#/c/13820/2/docs/installation.adoc@654 PS2, Line 654: kudu-master "the masters" http://gerrit.cloudera.org:8080/#/c/13820/2/docs/installation.adoc@655 PS2, Line 655: . Restore the default gflag value (5 minutes) for every kudu-tserver. : + : [source,bash] : ---- : ./kudu tserver set_flag <tserver> follower_unavailable_considered_failed_sec 300 --force : ---- Is this actually necessary? Won't restarting the process restore the default value of all of its gflags? Come to think of it, how does the gflag value overriding work if restarting a tserver restores the original value? Seems like we'd need to restart the tserver AND THEN override follower_unavailable_considered_failed_sec. -- To view, visit http://gerrit.cloudera.org:8080/13820 To unsubscribe, visit http://gerrit.cloudera.org:8080/settings Gerrit-Project: kudu Gerrit-Branch: master Gerrit-MessageType: comment Gerrit-Change-Id: I6b3e5c549dc05c3388c0b0dd628d205a356da344 Gerrit-Change-Number: 13820 Gerrit-PatchSet: 2 Gerrit-Owner: helifu <[email protected]> Gerrit-Reviewer: Adar Dembo <[email protected]> Gerrit-Reviewer: Kudu Jenkins (120) Gerrit-Reviewer: Priyanka Chheda <[email protected]> Gerrit-Reviewer: helifu <[email protected]> Gerrit-Comment-Date: Wed, 10 Jul 2019 14:19:13 +0000 Gerrit-HasComments: Yes
