[
https://issues.apache.org/jira/browse/HBASE-18070?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17167629#comment-17167629
]
Andrew Kyle Purtell commented on HBASE-18070:
---------------------------------------------
Meta replicas can be safely out of date. Stale data in a replica just means
region relocation may fail, the client will get a NSRE. If then the client
falls back to read from the primary, it will get the up to date location.
Generally clients will handle out of date replicas well, because their use of
meta is read only, and already handle out of date caches. You can think of out
of date local region location cache and stale meta replica as equivalent. Of
course they have to be able to ask again of the primary or else they’ll get
stuck. We need to ensure this happens.
What would be a problem is if something, like the master, reads from a
possibly out of date replica and then mutates cluster state based on that
information. Presumably the master’s components like AssignmentManager and
ServerManager will refuse to talk to replicas. We also have to make sure this
is the case.
Then, meta read replicas should be fine even if stale and even if updates come
in out of order. Ideally we fix the ordering though. The Casual Replication
issue frames this problem and a solution for cross cluster replication and
should apply to the meta read replica case as well.
> Enable memstore replication for meta replica
> --------------------------------------------
>
> Key: HBASE-18070
> URL: https://issues.apache.org/jira/browse/HBASE-18070
> Project: HBase
> Issue Type: New Feature
> Reporter: Hua Xiang
> Assignee: Huaxiang Sun
> Priority: Major
>
> Based on the current doc, memstore replication is not enabled for meta
> replica. Memstore replication will be a good improvement for meta replica.
> Create jira to track this effort (feasibility, design, implementation, etc).
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
