[
https://issues.apache.org/jira/browse/HBASE-28342?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Nick Dimiduk updated HBASE-28342:
---------------------------------
Release Note:
<!-- markdown -->
This change introduces the configuration
`hbase.master.reject.decommissioned.hosts`. When this property is set to
`true`, region servers added to the [decommissioning hosts
list](https://hbase.apache.org/apidocs/org/apache/hadoop/hbase/client/Admin.html#decommissionRegionServers-java.util.List-boolean-)
will be checked by hostname only (not taking into consideration RPC port or
startcode). When a region server with a hostname that matches the list attempts
to join the cluster, the Master will reject its application by responding with
the new `DecommissionedHostRejectedException`.
> Decommissioned hosts should be rejected by the HMaster
> ------------------------------------------------------
>
> Key: HBASE-28342
> URL: https://issues.apache.org/jira/browse/HBASE-28342
> Project: HBase
> Issue Type: Improvement
> Components: master
> Reporter: Ahmad Alhour
> Assignee: Ahmad Alhour
> Priority: Major
> Labels: pull-request-available
> Fix For: 2.6.0, 4.0.0-alpha-1, 3.0.0-beta-2
>
>
> We had an issue with a cluster, internally at HubSpot, where a decommissioned
> RegionServer was still being picked up by the HMaster. The host the
> RegionServer was living on was impaired, and we couldn't correctly kill the
> RegionServer, so the HMaster would periodically hear back from the host and
> remove it from its dead host's list.
> We would like to implement a fix so that this doesn't happen. We're thinking
> of adding a boolean flag to the Decommission RegionServer Admin API that
> signifies ignoring the startcode of the servername, when the boolean is True
> the host will be rejected every time it comes back even if it had a different
> startcode.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
