From 34755498b822c532d63c01c4f7bd578e57442023 Mon Sep 17 00:00:00 2001 From: Abhishek Lekshmanan Date: Fri, 8 Mar 2019 16:57:28 +0100 Subject: [PATCH] doc: add troubleshooting notes on reshard admin clis Adding a note on LC fixes and reshard stale instance fixes Signed-off-by: Abhishek Lekshmanan (cherry picked from commit dee9ac22f19360ba6436e674c93baea9d97ca5da) --- doc/radosgw/dynamicresharding.rst | 42 +++++++++++++++++++++++++++++++ 1 file changed, 42 insertions(+) diff --git a/doc/radosgw/dynamicresharding.rst b/doc/radosgw/dynamicresharding.rst index 4d51cd76ebc5c..cd1ebdce808d5 100644 --- a/doc/radosgw/dynamicresharding.rst +++ b/doc/radosgw/dynamicresharding.rst @@ -95,3 +95,45 @@ Manual bucket resharding :: # radosgw-admin bucket reshard --bucket --num-shards + + +Troubleshooting +=============== + +Clusters prior to Luminous 12.2.11 and Mimic 13.2.5 left behind stale bucket +instance entries that weren't automatically cleaned up. The issue also affected +LifeCycle policies which weren't applied to resharded buckets anymore. Both of +these issues can be worked around using a couple of radosgw-admin commands. + +Stale Instance Management +------------------------- + +:: + + # radosgw-admin reshard stale-instances list + +This lists the stale instances in a cluster that are ready to be cleaned up. +Please note that the cleanup of these instances should be done only on a single +site cluster. The cleanup can be done by the following command: + +:: + + # radosgw-admin reshard stale-instances rm + + +Lifecycle fixes +--------------- + +For clusters which had resharded instances, it is highly likely that the old +lifecycle processes would've flagged and deleted lifecycle processing as the +bucket instance changed during a reshard. While this is fixed for newer clusters +(from 13.2.6 and 12.2.12), older buckets which had lifecycle policies and +would've undergone reshard will have to be manually fixed by issuing the following command + +:: + + # radosgw-admin lc reshard fix --bucket {bucketname} + + +As a convenience wrapper, if the ``--bucket`` argument is dropped then this +command will try and fix LC policies for all the buckets in the cluster. -- 2.39.5