[ec2] Can't contact instance

[Tom Haddon]

On Tue, 2009-08-04 at 20:26 -0700, Eric Hammond wrote:
> Tom:
> 
> Almost everything I know about building Ubuntu AMIs is in code form in
> http://ec2ubuntu-build-ami.notlong.com
> 
> There should be a few minor changes if you are running on the Canonical
> kernels (e.g., kernel modules) but if you go through this code something
> there might trigger an idea to test.
> 
> You don't need to set up /etc/hostname or set any hostname in /etc/hosts
> other than the usual "localhost".  DHCP can set the regular hostname for
> you using info from Amazon.

I'm clearly a muppet. The problem was an
incomplete /etc/network/interfaces. Everything seems to be working fine
now.

Thanks, Tom

> 
> --
> Eric Hammond
> ehammond at thinksome.com
> 
> 
> 
> Tom Haddon wrote:
> > On Tue, 2009-08-04 at 15:23 -0700, Eric Hammond wrote:
> >> Tom:
> >>
> >> I had, perhaps incorrectly, assumed you built your AMI by rebundling a
> >> running instance of a public AMI.
> >>
> >> Did you build your AMI from scratch using vmbuilder,
> >> ec2ubuntu-build-ami, or by hand?
> > 
> > I built it from scratch. 
> > 
> >> There are various EC2 specific boot tasks that need to be taken care of
> >> without which you cannot connect to the instance.
> > 
> > Yeah, the ones that I'd been doing were working fine on my i386 images,
> > just things didn't seem to be working for this amd64 image (the first
> > I'd created). Basically my process was:
> > 
> > - Create a sparse file
> > - Mount the sparse file as a loopback device (let's say it's mounted at
> > dir_name)
> > - debootstrap --arch amd64 hardy dir_name
> > - In the chroot:
> >  - add /sbin/ldconfig to /etc/rc.local
> >  - create /etc/ld.so.conf.d/xen.conf with: hwcap 0 nosegneg
> >  - add /etc/init.d/ec2-ssh-host-key and run update-rc.d per
> > http://paste.ubuntu.com/247568/
> >  - add my own user accounts/ssh authentication/sudo setup
> >  - any other setup/packages you want for the particular AMI
> > - bundle and upload to AWS
> > 
> > I think there may have been a mismatch between /etc/hosts
> > and /etc/hostname - I've fixed that and will be re-uploading the image
> > and trying again.
> > 
> > Thanks, Tom
> > 
> >> --
> >> Eric Hammond
> >> ehammond at thinksome.com
> >>
> >>
> >>
> >> Tom Haddon wrote:
> >>> On Mon, 2009-08-03 at 19:44 -0700, Eric Hammond wrote:
> >>>> Tom:
> >>>>
> >>>> Right about the time you reported this, a higher than usual number of
> >>>> other people reported similar problems on the EC2 forum.  There may have
> >>>> been a localized networking or systems issue.
> >>>>
> >>>> Amazon's AWS status board has not shown any problems at this point, so
> >>>> it may have been temporary or isolated.
> >>>>
> >>>> I would recommend trying to start new instances, perhaps in multiple
> >>>> availability zones, to see if the problem is really with your AMI or the
> >>>> environment.
> >>> I tried starting instances of the AMI in all four availability zones and
> >>> none of the instances were contactable. I'm going to try comparing my
> >>> AMI with the stock ubuntu amd64 hardy AMI and see if I can see any major
> >>> differences.
> >>>
> >>> Thanks, Tom
> >>>
> >>>> --
> >>>> Eric Hammond
> >>>> ehammond at thinksome.com
> >>>>
> >>>>
> >>>>
> >>>> Tom Haddon wrote:
> >>>>> Hi Folks,
> >>>>>
> >>>>> I'm getting a very strange issue with a particular AMI I've created.
> >>>>> It's a x86_64 image, and I'm using the stock Ubuntu AKI/ARI
> >>>>> (aki-6f709706/ari-61709708). I start an instance of my AMI and get the
> >>>>> following console output:
> >>>>>
> >>>>> http://paste.ubuntu.com/246739/
> >>>>>
> >>>>> I can't login to the instance, or contact it via ping, even though:
> >>>>>
> >>>>> http://paste.ubuntu.com/246741/
> >>>>>
> >>>>> Any ideas what could be going on? From what I can see of the console
> >>>>> output the machine has booted okay, but then I just can't contact it at
> >>>>> all...
> >>>>>
> >>>>> Thanks, Tom
> >>>>>
> >>>>>
> >