Demai Ni 2013-08-16, 16:58
Ted Yu 2013-08-16, 17:07
Demai Ni 2013-08-16, 17:32
Ted Yu 2013-08-16, 17:38
Demai Ni 2013-08-16, 17:52
Ted Yu 2013-08-16, 17:53
Ted Yu 2013-08-17, 01:35
Demai Ni 2013-08-17, 03:09
Ted Yu 2013-08-18, 03:20
Stack 2013-08-18, 16:35
Demai Ni 2013-08-18, 18:37
Stack 2013-08-19, 18:09
yeah. the tool will be applied when hbase on master cluster is down. In the
case that user decides not to restart hbase(for whatever reason), the tool
will be used to push the pending Edits over to slave peers.
I will look into RecoverableZooKeeper. thanks for the suggestion
On Mon, Aug 19, 2013 at 11:09 AM, Stack <[EMAIL PROTECTED]> wrote:
> On Sun, Aug 18, 2013 at 11:37 AM, Demai Ni <[EMAIL PROTECTED]> wrote:
> > St.Ack and Ted,
> > many thanks for your help and suggestions on this question. I am working
> > HBASE-9047 while encountering this question. To provide a whole picture
> > what I am looking for, I uploaded the prototype code and some background
> > info over there(
> > https://issues.apache.org/jira/browse/HBASE-9047#comment-13743322).
> > Basically, I am looking for a way to get the original zookeeper when
> > is offline, so that a new replicationManager can be created and
> > the pending replication Edits, once get them, the replicationManger will
> > its magic using the logic as if hbase is restarting. will
> > RecoverableZooKeeper instance have the access to the replication info
> > original zookeeper?
> hbase is down at this time?
> The replication state is kept out in zk so yeah, any connection to zk will
> have a view on current state of replication.