[
https://issues.apache.org/jira/browse/HBASE-19394?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16289577#comment-16289577
]
Josh Elser commented on HBASE-19394:
------------------------------------
bq. But I guess they wanted to reduce a downtime when RS goes down by using
this facility
That's the best thing I could extrapolate. It's a little silly because noticing
(at the transport layer) that a RS is dead is not remotely close to being the
blocker in getting access to that data again (WAL recovery is going to take
many seconds whereas the RPC layer should. realize that a remote host is "down"
within 10s to 100s of millis).
> Support multi-homing env for the publication of RS status with multicast
> (hbase.status.published)
> --------------------------------------------------------------------------------------------------
>
> Key: HBASE-19394
> URL: https://issues.apache.org/jira/browse/HBASE-19394
> Project: HBase
> Issue Type: Bug
> Components: Client, master
> Reporter: Toshihiro Suzuki
> Assignee: Toshihiro Suzuki
> Fix For: 2.0.0-beta-1
>
> Attachments: HBASE-19394.patch
>
>
> Currently, when the publication feature is enabled
> (hbase.status.published=true), it uses the interface which is found first:
> https://github.com/apache/hbase/blob/2e8bd0036dbdf3a99786e5531495d8d4cb51b86c/hbase-server/src/main/java/org/apache/hadoop/hbase/master/ClusterStatusPublisher.java#L268-L275
> This won't work when the host has the multiple network interfaces and the
> unreachable one to the other nodes is selected. The interface which can be
> used for the communication between cluster nodes should be configurable.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
