[
https://issues.apache.org/jira/browse/HBASE-12035?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14141892#comment-14141892
]
Mikhail Antonov commented on HBASE-12035:
-----------------------------------------
bq. It is also wrong having client going to master to figure is 'table is
enabled/disabled'.
[~stack] - is that deemed wrong because it's expensive (vs. going to ZK, or vs.
not going anywhere at all?), or because it puts additional requirement on
master to "always" be up?
If latter, then do we want to separately think about the cases when meta is
co-located, and when it's not? In it is co-located, then this additional
requirement for HA should be already present (if master is down, meta is down
too, and whole region assignment/lookup doesn't work)? I'm thinking in the
following direction. If meta isn't co-located, then probability of both meta RS
and master going down is much lower than probability of co-located master+meta
going down, but probability of either one going down is higher. On the other
hand, when either master or meta in non-colocated setup is down, cluster is
harmed somewhat.
I guess I'm curious if folks are avoiding co-located setup for that reason?
Pinging [~eclark].
> Client does an RPC to master everytime a region is relocated
> ------------------------------------------------------------
>
> Key: HBASE-12035
> URL: https://issues.apache.org/jira/browse/HBASE-12035
> Project: HBase
> Issue Type: Improvement
> Reporter: Enis Soztutar
> Priority: Critical
> Fix For: 2.0.0
>
>
> HBASE-7767 moved table enabled|disabled state to be kept in hdfs instead of
> zookeeper. isTableDisabled() which is used in
> HConnectionImplementation.relocateRegion() now became a master RPC call
> rather than a zookeeper client call. Since we do relocateRegion() calls
> everytime we want to relocate a region (region moved, RS down, etc) this
> implies that when the master is down, the some of the clients for uncached
> regions will be affected.
> See HBASE-7767 and HBASE-11974 for some more background.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
