hbase-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Mark Kerzner <mark.kerz...@shmsoft.com>
Subject Re: Do I really need fully-distributed mode for HBase?
Date Fri, 30 Dec 2011 22:17:25 GMT
I found the answer, here it is for everyone's benefit,

If you want to access HBase from a remote client (like any hadoop node, for
example), then you run into this problem: HBase master auto-discovers
itself as localhost (using ‘hostname’ command, and tells this to zookeeper,
which tells this to your client in turn, and it can’t connect. See here,

And the answer is, tell the HBase master who he is, using the internal ip
of EC2 (did I mention that I am on EC2?), in /etc/hosts

#      domU-12-31-38-04-AA-53.compute-1.internal
domU-12-31-38-04-AA-53  domU-12-31-38-04-AA-53.compute-1.internal

On Thu, Dec 29, 2011 at 11:37 PM, Mark Kerzner <markkerzner@gmail.com>wrote:

> Hi,
> if I am running HBase in pseudo-distributed mode and I cannot connect to it
> from other nodes, I guess, that is to be expected, right?
> I mean, the browser connects to http://hbase-url:60010, but the zookeeper,
> which ignored my ip and started on localhost, would give me a problem. So I
> would need 3 nodes for zookeeper. And that's the reason for me error
> 11/12/30 04:48:56 INFO zookeeper.ClientCnxn: Session establishment complete
> on server domU-12-31-38-04-AA-53.compute-1.internal/,
> sessionid = 0x1348d4c23a60005, negotiated timeout = 40000
> 11/12/30 04:48:56 INFO ipc.HbaseRPC: Server at
> localhost/ not be reached after 1 tries, giving
> up.
> Thank you for any pointers.
> Regards,
> Mark

  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message