[
https://issues.apache.org/jira/browse/HBASE-30220?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18088081#comment-18088081
]
Kevin Geiszler commented on HBASE-30220:
----------------------------------------
This has been discussed before. Although it is easy to throw an error after
the {{update_all_config}} command and stop the update on master, we also need
to figure out how to prevent the region servers from executing with their
respective {{onConfigurationChange()}} methods and disabling read-only mode on
their own.
https://github.com/apache/hbase/pull/7922/changes#r2932155128
> A replica cluster can have read-only mode disabled even when another active
> cluster already exists
> --------------------------------------------------------------------------------------------------
>
> Key: HBASE-30220
> URL: https://issues.apache.org/jira/browse/HBASE-30220
> Project: HBase
> Issue Type: Bug
> Components: read replicas
> Reporter: Kevin Geiszler
> Assignee: Kevin Geiszler
> Priority: Major
>
> Assume you have a Read-Replica cluster setup where one cluster is the Active
> Cluster and the other cluster is a Replica Cluster with read-only mode
> enabled. Currently, it is possible to disable read-only mode on the Replica
> Cluster and end up with two clusters that allow writes. This should not
> happen.
> The steps to reproduce this are simple:
> # Start one cluster as the Active Cluster and another as the Replica Cluster
> # You should be able to create a table and add data on the Active Cluster
> but not on the Replica Cluster.
> # Change the {{hbase.global.readonly.enabled}} property from {{true}} to
> {{false}} on the Replica, save the conf file, and run
> {{{{update_all_config}}}}
> # Verify you can now create a table and add data on both clusters
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
