On Mon, Jun 21, 2010 at 10:47 AM, Oliver Meyn <[email protected]
wrote:
Thanks for your response - I agree that this looks
like an ip problem.
Further digging suggests I'm lost in the cracks
that patch 1279 tries to
solve (https://issues.apache.org/jira/browse/HBASE-1279).
I think what
I'm missing is a working example of how the new config
should work.
Does anyone with an ec2 cluster (or test machine)
that's accessible to
clients from the outside world care to share their
config? Of most interest
are the hbase-site.xml entries for *.dns.interface,
*.dns.nameserver, and
hbase.zookeper.quorom. Naturally the whole file
would be most illuminating.
Thanks,
Oliver
On 19-Jun-10, at 1:30 AM, 豊月 wrote:
hi Oliver.
I found this problem too.
this probrem came from EC2's system & perhaps
Today's HBase can't use that
case.
your request from client try connect to hbase by
External DNS/IP.
when request reachs EC2, EC2 change Externl IP to
EC2's Internal IP.
HBase get request, then return Internal IP which
has requested Data is.
so your Client try connect with EC2's innternal
IP.
Client & you Expect External IP. but HBase
Return Internal IP.
thats the why your Client can't get data.
sorry i'm poor English
thanks.
2010/6/19 Oliver Meyn <[email protected]>:
Hi All,
Is there a trick to running a single,
standalone hbase on a single ec2
instance? I have hbase running locally
as standalone and again on a
separate testing machine in the office and my
java client can talk to
them
both just fine. If I setup an ec2
instance with the same configs as the
testing machine my client can't connect -
meaning it attempts to connect
and
then hangs with no further logging. If I
turn off hbase on the ec2
instance
and try to connect I get the usual
ConnectionException: Connection
refused.
When hbase is running I can connect to the
admin console on :60010 just
fine, and I've confirmed that there are no
blocked ports facing me.
I suspect this has something to with ip
addresses, specifically the
elastic
ip story of the external ip not being known to
the instance (where it
instead has a 10.x.x.x address), but fiddling
hasn't helped.
Any thoughts?
Thanks,
Oliver