HBASE-2849 - 90.3 hbase clients recovering from invalid zk sessions

[Douglas Campbell <de...@yahoo.com>]

Did this patch make it into 0.90.3?

https://issues.apache.org/jira/browse/HBASE-2849

When I go into HConnectionManager source from the 90.3 branch,  evidence of the patch is missing.

Pointers to where this logic is would be helpful.

I'm trying to make sure my clients reconnect under zk and hbase failure scenarios.  Any information surrounding that would also be appreciated.


-Doug

Re: HBASE-2849 - 90.3 hbase clients recovering from invalid zk sessions

[Jean-Daniel Cryans <jd...@apache.org>]

That code was moved inside ZooKeeperWatcher during the master rewrite:

      case Expired:
        String msg = prefix(this.identifier + " received expired from " +
          "ZooKeeper, aborting");
        // TODO: One thought is to add call to ZooKeeperListener so say,
        // ZooKeperNodeTracker can zero out its data values.
        if (this.abortable != null) this.abortable.abort(msg,
            new KeeperException.SessionExpiredException());
        break;

Then in HConnectionManager this is handled like that in abort():

          LOG.info("This client just lost it's session with ZooKeeper, trying" +
              " to reconnect.");
          resetZooKeeperTrackers();
          LOG.info("Reconnected successfully. This disconnect could have been" +
              " caused by a network partition or a long-running GC pause," +
              " either way it's recommended that you verify your environment.");

Hope this helps,

J-D

On Wed, Aug 3, 2011 at 9:43 AM, Douglas Campbell <de...@yahoo.com> wrote:
> Did this patch make it into 0.90.3?
>
> https://issues.apache.org/jira/browse/HBASE-2849
>
> When I go into HConnectionManager source from the 90.3 branch,  evidence of the patch is missing.
>
> Pointers to where this logic is would be helpful.
>
> I'm trying to make sure my clients reconnect under zk and hbase failure scenarios.  Any information surrounding that would also be appreciated.
>
>
> -Doug