custom (ebs backed) Fedora 14 AMIs

[Brian LaMere]

Greetings all - just wanted to let you know I made a couple tweeks to the
very very simple, very non-professional, non-bulletproof
"createfedorabootebs.sh" script.  You can find slightly longer instructions
here:

https://fedoraproject.org/wiki/Publishing_image_to_EC2

The steps are relatively simple.

1)  create an instance from the official AMI (or anywhere else, really)
2)  create a new volume, whatever size (at least 2g or so) and mount it at
/dev/sdf
3)  run the createfedorabootebs.sh script
4)  snapshot the volume
5)  register an AMI with the snapshot and the pvgrub hd0 aki

It's a simple script; even a novice can tell what is going on.  The sources
are all from Fedora, and nothing funny is happening.   This is the
low-speed, "I don't know boxgrinder, I don't have a machine I can run xen on
anyway, I want to know what is on the image" approach.

That said, I *heartily* recommend using jforbes' AMIs instead; he'll have
EBS-backed AMIs soon, I'm sure.  This is only for those who want to play, or
need an ebs image today.  Or, if you're really crazy, it's for people who
want to look at getting rawhide on ec2!

Brian

[Marek Goldmann]

Hi Brian,

Good writeup!

From my side I propose to use BoxGrinder, and you'll have 2 simple steps:

1. Create appliance definition file (f13-jeos.appl):

name: f13-jeos
summary: Just Enough Operating System based on Fedora 13
os:
  name: fedora
  version: 13
hardware:
  partitions:
    "/":
      size: 2
packages:
  includes:
    - @core

2. Run BoxGrinder:

boxgrinder-build f13-jeos.appl -p ec2 -d ebs

That's all. It'll build the appliance, convert to EC2 format, create volume, make snapshot, rsync files and finally register the snapshot as AMI. Of course you need to run it on an EC2 instance to have it available. You can use our meta-appliances to do the build: http://www.jboss.org/boxgrinder/downloads/build/meta-appliance .

HTH

--Marek

On 2010-12-04, at 01:28, Brian LaMere wrote:
> Greetings all - just wanted to let you know I made a couple tweeks to the very very simple, very non-professional, non-bulletproof "createfedorabootebs.sh" script.  You can find slightly longer instructions here:
>
> https://fedoraproject.org/wiki/Publishing_image_to_EC2
>
> The steps are relatively simple.
> create an instance from the official AMI (or anywhere else, really)
> create a new volume, whatever size (at least 2g or so) and mount it at /dev/sdf
> run the createfedorabootebs.sh script
> snapshot the volume
> register an AMI with the snapshot and the pvgrub hd0 aki
>
> It's a simple script; even a novice can tell what is going on.  The sources are all from Fedora, and nothing funny is happening.   This is the low-speed, "I don't know boxgrinder, I don't have a machine I can run xen on anyway, I want to know what is on the image" approach.
>
> That said, I *heartily* recommend using jforbes' AMIs instead; he'll have EBS-backed AMIs soon, I'm sure.  This is only for those who want to play, or need an ebs image today.  Or, if you're really crazy, it's for people who want to look at getting rawhide on ec2!
>
> Brian
> _______________________________________________
> cloud mailing list
> [e-mail]
> https://admin.fedoraproject.org/mailman/listinfo/cloud

[John Poelstra]

Brian LaMere said the following on 12/03/2010 04:28 PM Pacific Time:
> Greetings all - just wanted to let you know I made a couple tweeks to
> the very very simple, very non-professional, non-bulletproof
> "createfedorabootebs.sh" script.  You can find slightly longer
> instructions here:
>
> https://fedoraproject.org/wiki/Publishing_image_to_EC2
>
> The steps are relatively simple.
> create an instance from the official AMI (or anywhere else, really)
> create a new volume, whatever size (at least 2g or so) and mount it
>
> at /dev/sdf
> 3)  run the createfedorabootebs.sh script
> 4)  snapshot the volume
> 5)  register an AMI with the snapshot and the pvgrub hd0 aki
>
> It's a simple script; even a novice can tell what is going on.  The
> sources are all from Fedora, and nothing funny is happening.   This is
> the low-speed, "I don't know boxgrinder, I don't have a machine I can
> run xen on anyway, I want to know what is on the image" approach.
>
> That said, I *heartily* recommend using jforbes' AMIs instead; he'll
> have EBS-backed AMIs soon, I'm sure.  This is only for those who want to
> play, or need an ebs image today.  Or, if you're really crazy, it's for
> people who want to look at getting rawhide on ec2!
>
> Brian

What are the plans and timeline for moving this function to the Fedora 
Release Engineering team?

All "official Fedora content" should be produced and staged by Release 
Engineering, and tested before release by QA too.

John

[Brian LaMere]

The script isn't intending to be "official" - jforbes' AMIs fulfill that
role.  This is just a working how-to, for those wanting basic starting tips
on creating their own fedora AMI.  It's actually not worthwhile to spend
terribly much effort making a good tool out of it; just use Boxgrinder or
such if you want a real tool ;)  I believe boxgrinder is moving along the
"official" path already.

Currently (and temporarily) the published AMIs are only S3-backed; some of
us need EBS-backed instances, so for those who do I reminded everyone of
that howto, and let them know I re-verified it works.  I have a happy,
never-ran-prior-to-use, AMI for Fedora14 with an EBS the size I declare.

Brian

On Mon, Dec 6, 2010 at 10:35 AM, John Poelstra  wrote:
> What are the plans and timeline for moving this function to the Fedora
> Release Engineering team?
>
> All "official Fedora content" should be produced and staged by Release
> Engineering, and tested before release by QA too.
>
> John
> _______________________________________________
> cloud mailing list
> [e-mail]
> https://admin.fedoraproject.org/mailman/listinfo/cloud

[John Poelstra]

Brian LaMere said the following on 12/06/2010 11:10 AM Pacific Time:
> The script isn't intending to be "official" - jforbes' AMIs fulfill that
> role.  This is just a working how-to, for those wanting basic starting
> tips on creating their own fedora AMI.  It's actually not worthwhile to
> spend terribly much effort making a good tool out of it; just use
> Boxgrinder or such if you want a real tool ;)  I believe boxgrinder is
> moving along the "official" path already.
>
> Currently (and temporarily) the published AMIs are only S3-backed; some
> of us need EBS-backed instances, so for those who do I reminded everyone
> of that howto, and let them know I re-verified it works.  I have a
> happy, never-ran-prior-to-use, AMI for Fedora14 with an EBS the size I
> declare.

My point is that going forward "jforbes' AMIs" should be produced by 
Release Engineering not that you're doing something wrong.

John

[Robyn Bergeron]

On Mon, Dec 6, 2010 at 11:35 AM, John Poelstra  wrote:
> Brian LaMere said the following on 12/03/2010 04:28 PM Pacific Time:
> > Greetings all - just wanted to let you know I made a couple tweeks to
> > the very very simple, very non-professional, non-bulletproof
> > "createfedorabootebs.sh" script.  You can find slightly longer
> > instructions here:
> >
> > https://fedoraproject.org/wiki/Publishing_image_to_EC2
> >
> > The steps are relatively simple.
> > create an instance from the official AMI (or anywhere else, really)
> > create a new volume, whatever size (at least 2g or so) and mount it
> >
> > at /dev/sdf
> > 3)  run the createfedorabootebs.sh script
> > 4)  snapshot the volume
> > 5)  register an AMI with the snapshot and the pvgrub hd0 aki
> >
> > It's a simple script; even a novice can tell what is going on.  The
> > sources are all from Fedora, and nothing funny is happening.   This is
> > the low-speed, "I don't know boxgrinder, I don't have a machine I can
> > run xen on anyway, I want to know what is on the image" approach.
> >
> > That said, I *heartily* recommend using jforbes' AMIs instead; he'll
> > have EBS-backed AMIs soon, I'm sure.  This is only for those who want to
> > play, or need an ebs image today.  Or, if you're really crazy, it's for
> > people who want to look at getting rawhide on ec2!
> >
> > Brian
>
> What are the plans and timeline for moving this function to the Fedora
> Release Engineering team?
>
> All "official Fedora content" should be produced and staged by Release
> Engineering, and tested before release by QA too.
>
> John

We discussed this briefly at the cloud SIG meeting last week.  Jared
proposed that we do a FAD around this, probably within a few weeks of
FUDCon.  We can likely use FUDCon to hash out some of the details, and
then hack out the implementation, plus docs and test plans, at the
FAD.  Jared was planning on mailing the list this week with some
initial thoughts.

Allow me to play devil's advocate for a moment here: We are, in
essence, doing for Amazon, what some other cloud providers do for us.
Fedora 14 on Rackspace's Cloud Servers and Slicehost products, for
example,  required no intervention, or even asking from the Cloud SIG
- they just put it up.  Where do we draw the line on what should be
produced and staged by rel-eng? Is it simply because *we* are the ones
doing the heavy lifting with EC2, and not Amazon?

-robyn
> cloud mailing list
> [e-mail]
> https://admin.fedoraproject.org/mailman/listinfo/cloud

[Steve Gordon]

On 12/07/2010 06:20 AM, Robyn Bergeron wrote:
> On Mon, Dec 6, 2010 at 11:35 AM, John Poelstra  wrote:
> > Brian LaMere said the following on 12/03/2010 04:28 PM Pacific Time:
> > > Greetings all - just wanted to let you know I made a couple tweeks to
> > > the very very simple, very non-professional, non-bulletproof
> > > "createfedorabootebs.sh" script.  You can find slightly longer
> > > instructions here:
> > >
> > > https://fedoraproject.org/wiki/Publishing_image_to_EC2
> > >
> > > The steps are relatively simple.
> > > create an instance from the official AMI (or anywhere else, really)
> > > create a new volume, whatever size (at least 2g or so) and mount it
> > >
> > > at /dev/sdf
> > > 3)  run the createfedorabootebs.sh script
> > > 4)  snapshot the volume
> > > 5)  register an AMI with the snapshot and the pvgrub hd0 aki
> > >
> > > It's a simple script; even a novice can tell what is going on.  The
> > > sources are all from Fedora, and nothing funny is happening.   This is
> > > the low-speed, "I don't know boxgrinder, I don't have a machine I can
> > > run xen on anyway, I want to know what is on the image" approach.
> > >
> > > That said, I *heartily* recommend using jforbes' AMIs instead; he'll
> > > have EBS-backed AMIs soon, I'm sure.  This is only for those who want to
> > > play, or need an ebs image today.  Or, if you're really crazy, it's for
> > > people who want to look at getting rawhide on ec2!
> > >
> > > Brian
> >
> > What are the plans and timeline for moving this function to the Fedora
> > Release Engineering team?
> >
> > All "official Fedora content" should be produced and staged by Release
> > Engineering, and tested before release by QA too.
> >
> > John
>
> We discussed this briefly at the cloud SIG meeting last week.  Jared
> proposed that we do a FAD around this, probably within a few weeks of
> FUDCon.  We can likely use FUDCon to hash out some of the details, and
> then hack out the implementation, plus docs and test plans, at the
> FAD.  Jared was planning on mailing the list this week with some
> initial thoughts.
>
> Allow me to play devil's advocate for a moment here: We are, in
> essence, doing for Amazon, what some other cloud providers do for us.
> Fedora 14 on Rackspace's Cloud Servers and Slicehost products, for
> example,  required no intervention, or even asking from the Cloud SIG
> they just put it up.  Where do we draw the line on what should be
>
> produced and staged by rel-eng? Is it simply because *we* are the ones
> doing the heavy lifting with EC2, and not Amazon?

The AMIs were noted in the release notes and marketing material around 
the release as a feature of the F14 release. To me they're effectively 
being touted as an official release media and as such should be subject 
to the same processes as the DVD or LiveCD images.

While I agree that it's great that Rackspace and Slicehost have been so 
quick to get images up it's important to remember, at least in the case 
of Rackspace, this is primarily down to the goodwill of one of their 
employees who happens to be a Fedora enthusiast. Unfortunately we can't 
rely on this being the case at every cloud provider :(.

I would however think in future we might like to make some more noise 
about the presence of images we are aware of on other providers - 
obviously this would be at odds with my first statement about subjecting 
images to the normal rel-eng process. This is where we have a difficult 
balance to strike as some providers may not even make it an option for 
us to supply our own image.

The reason Amazon are able to take such a hands-off approach is that 
they are seen as the biggest player in the public cloud market (rightly 
or wrongly) and it feeds into that if they are the only provider we 
market ourselves as available on.

FWIW I am aware I am still just shooting from the peanut gallery here, I 
do still hope to be able to contribute to the Cloud Guide in the future.
> -robyn
> > cloud mailing list
> > [e-mail]
> > https://admin.fedoraproject.org/mailman/listinfo/cloud
>
> cloud mailing list
> [e-mail]
> https://admin.fedoraproject.org/mailman/listinfo/cloud