Hi All,

OK please see below for some progress on this one. We most certainly need to configure the slave node to build Nutch before we can get the nightly builds back online.

In the meantime, AFAIK the patches pending for commit have been tested and no objections have been raised.

I'll keep working on this.

Best

Lewis

---------- Forwarded message ----------
From: Andy Isaacson <adi@cloudera.com>
Date: Wed, Jan 9, 2013 at 12:56 PM
Subject: Re: UnknownHostException after upgrade from 1.0.3 > 1.1.1
To: user@hadoop.apache.org


On Tue, Jan 8, 2013 at 7:37 PM, Lewis John Mcgibbney
<lewis.mcgibbney@gmail.com> wrote:
> Over @Nutch, we are experiencing failed builds on our nightly CI builds
> after the above upgrade.
> This happens on the Jenkins (Solaris and Ubuntu) builds but NOT on local
> operating systems. This has been verified by numerous members of the
> community.
> The stack we get is appended below.
[snip]
> java.net.UnknownHostException: -s: -s

I guess this is an error from the Solaris build, and that the Linux
one is slightly different, with a different string in place of the
"-s"? The "-s" output implies that some Linux-specific script was run
as root on your Solaris machine and did a "hostname -s", which on
Linux prints the short hostname, but on Solaris sets the current
hostname of the machine to "-s". A reboot will clear up that problem,
or you can manually fix the hostname with "hostname
whatever-the-hostname-is". If you're doing your Jenkins builds as
root, I highly recommend that you stop doing that.

In both cases you can fix the build failure by ensuring that the build
machine's hostname (run the "hostname" command) is present in the
/etc/hosts file, per

http://www.thegeekstuff.com/2012/02/hadoop-standalone-installation/

There are "better" ways to fix the problem by ensuring that your
network's DHCP and DNS configuration is correct, but /etc/hosts is the
quick and easy way.

-andy



--
Lewis