[
https://issues.apache.org/jira/browse/HBASE-28503?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Ahmad Alhour updated HBASE-28503:
---------------------------------
Description:
Last month we shipped a config feature to allow the HMaster to reject
decommissioned hosts (see:
[HBASE-28342|https://issues.apache.org/jira/browse/HBASE-28342]). After some
testing internally at HubSpot, we discovered that the HMaster loses the entry
of a decommissioned host from the draining ZNode when the RegionServer restarts
or becomes dead.
Our proposal for this fix would be to keep entries in the draining ZNode as
long as the HMaster is configured to reject decommissioned hosts.
was:
Last month we shipped a config feature to allow the HMaster to reject
decommissioned hosts. After some testing internally at HubSpot, we discovered
that the HMaster loses the entry of a decommissioned host from the draining
ZNode when the RegionServer restarts or becomes dead.
Our proposal for this fix would be to keep entries in the draining ZNode as
long as the HMaster is configured to reject decommissioned hosts.
> Keep entries in draining ZNode when HMaster is configured to reject
> decommissioned hosts
> ----------------------------------------------------------------------------------------
>
> Key: HBASE-28503
> URL: https://issues.apache.org/jira/browse/HBASE-28503
> Project: HBase
> Issue Type: Bug
> Reporter: Ahmad Alhour
> Priority: Major
>
> Last month we shipped a config feature to allow the HMaster to reject
> decommissioned hosts (see:
> [HBASE-28342|https://issues.apache.org/jira/browse/HBASE-28342]). After some
> testing internally at HubSpot, we discovered that the HMaster loses the entry
> of a decommissioned host from the draining ZNode when the RegionServer
> restarts or becomes dead.
> Our proposal for this fix would be to keep entries in the draining ZNode as
> long as the HMaster is configured to reject decommissioned hosts.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
