A new release of the Ubuntu Enterprise Cloud Images for stable Ubuntu release 10.10 LTS (Lucid Lynx) is available at [1]. Images are available for download or immediate use on EC2 via published AMI ids.
There are a few important updates since the 20100427.1 release [2]. * linux-image-2.6.32-305-ec2 [3] updated (2.6.32-305.9 -> 2.6.32-308.15) * landscape-client [4] updated (1.5.0.1-0ubuntu0.10.04.0 -> 1.5.4-0ubuntu0.10.04.0) * cloud-init [5] updated (0.5.10-0ubuntu1 -> 0.5.10-0ubuntu1.2)
If you are using a previous release, you should update in order to receive the kernel change. The kernel cannot be changed in an Ubuntu 10.10 based image by package upgrade.
This is the first image refresh of 10.04 and it includes many updates since the original release. In addition to reduced time for system upgrade on first boot, you may also want to upgrade your base image to receive the cloud-init or landscape-client updates as they affect the system's first boot before a system update can be done.
I was wondering about that, the page with the AMI links has 10.04 in
the title. I am not that up to date on the version strings but that
discrepancy registered even in my pico-second attention span :)
On Aug 27, 2:04 pm, Michael Wood <esiot...@gmail.com> wrote:
> On 27 August 2010 22:50, Scott Moser <smo...@ubuntu.com> wrote:> A new release of the Ubuntu Enterprise Cloud Images for stable Ubuntu
> > release 10.10 LTS (Lucid Lynx) is available at [1]. Images are available
On Fri, 27 Aug 2010, Scott Moser wrote: > A new release of the Ubuntu Enterprise Cloud Images for stable Ubuntu > release 10.10 LTS (Lucid Lynx) is available at [1]. Images are available > for download or immediate use on EC2 via published AMI ids.
Correction. The above should have said: 10.04 LTS (Lucid Lynx)
> On Fri, 27 Aug 2010, Scott Moser wrote:
> > A new release of the Ubuntu Enterprise Cloud Images for stable Ubuntu
> > release 10.10 LTS (Lucid Lynx) is available at [1]. Images are available
> > for download or immediate use on EC2 via published AMI ids.
> Correction. The above should have said:
> 10.04 LTS (Lucid Lynx)
It looks like you have a small bug in your script producing the files
in http://uec-images.ubuntu.com/query/lucid/server/ (actually it is
the same with Maverick and the older ones). architecture (amd64 or
i386) is not mentioned in the ec2 file name.
lucid server release 20100827 amd64 server/releases/lucid/
release-20100827/ubuntu-10.04-server-uec-amd64.tar.gz ubuntu-
lucid-10.04--server-20100827
lucid server release 20100827 i386 server/releases/lucid/
release-20100827/ubuntu-10.04-server-uec-i386.tar.gz ubuntu-
lucid-10.04--server-20100827
> It looks like you have a small bug in your script producing the files
> inhttp://uec-images.ubuntu.com/query/lucid/server/(actually it is
> the same with Maverick and the older ones). architecture (amd64 or
> i386) is not mentioned in the ec2 file name.
> It looks like you have a small bug in your script producing the files > in http://uec-images.ubuntu.com/query/lucid/server/ (actually it is > the same with Maverick and the older ones). architecture (amd64 or > i386) is not mentioned in the ec2 file name.
Thanks for catching this. I just recently added that field, and just now added a fix to the script.
On Fri, 2010-08-27 at 17:28 -0400, Scott Moser wrote: > On Fri, 27 Aug 2010, Scott Moser wrote:
> > A new release of the Ubuntu Enterprise Cloud Images for stable Ubuntu > > release 10.10 LTS (Lucid Lynx) is available at [1]. Images are available > > for download or immediate use on EC2 via published AMI ids.
> Correction. The above should have said: > 10.04 LTS (Lucid Lynx)
> A new release of the Ubuntu Enterprise Cloud Images for stable Ubuntu
> release 10.10 LTS (Lucid Lynx) is available at [1]. Images are available
> for download or immediate use on EC2 via published AMI ids.
> There are a few important updates since the 20100427.1 release [2].
> * linux-image-2.6.32-305-ec2 [3] updated (2.6.32-305.9 -> 2.6.32-308.15)
> * landscape-client [4] updated
> (1.5.0.1-0ubuntu0.10.04.0 -> 1.5.4-0ubuntu0.10.04.0)
> * cloud-init [5] updated (0.5.10-0ubuntu1 -> 0.5.10-0ubuntu1.2)
> If you are using a previous release, you should update in order to receive
> the kernel change. The kernel cannot be changed in an Ubuntu 10.10 based
> image by package upgrade.
> This is the first image refresh of 10.04 and it includes many updates
> since the original release. In addition to reduced time for system
> upgrade on first boot, you may also want to upgrade your base image to
> receive the cloud-init or landscape-client updates as they affect
> the system's first boot before a system update can be done.
On Sat, 28 Aug 2010, Darren Govoni wrote: > Scott, > Is it possible to perform a system update to the prior 10.04 AMI > release on a running instance (EBS) and get in sync that way?
It is possible to update an EBS instance. If you're running a previous ami , say us-east-1 ami-2d4aa444 ubuntu-lucid-10.04-i386-server-20100427.1 and you want to get to us-east-1 ami-1437dd7d ubuntu-lucid-10.04-i386-server-20100827
This will essentially be the same as 'apt-get update && apt-get dist-upgrade'.
Well, I was just trying to avoid duplication of data, and aiding people to get really only what they need/want rather than everything. Its not perfect, and the overhead of multiple requests is probably worse than one request for more data since we're dealing with pretty small files.
The usage that I was designing for was basically people polling http://uec-images.ubuntu.com/query/released.latest.txt , and determining if they were out of date based on that data. If so, they would do further queries to figure out what actions needed to be done.
That said:
http://uec-images.ubuntu.com/query/released.latest.txt gives you a row for each unique build in the released 'stream'. The 'released' stream contains builds labeled 'alpha-*', 'beta-*', and 'rc', and 'release'. The other stream is the 'daily' stream.
The primary key for a build is a combination of the 4 fields: suite, build_name, label, serial. ie: hardy server release 20100827
That file has 7 fields The first 4 of the key (suite, build_name, label, serial) Followed by arch, path, and 'suggested name'.
So, by walking each row in the released.latest.txt, and getting the appropriate released-dl.txt, you can get the data you want.
Ie, this will get you what you wanted: burl="http://uec-images.ubuntu.com/query"; str="released" wget -q "${burl}/${str}.latest.txt" -O - | while read suite bn lab ser; do wget -q -O - "${burl}/${suite}/${bn}/${str}-dl.txt" ; done
Alternatively, the data on uec-images is available via rsync. So, the following will also get you want you want: rsync -a uec-images.ubuntu.com::uec-images/query . cat query/*/*/released-dl.txt
Thank you for your answer. I had actually reached the solution of wget-
ing files for every single suite. I should have posted that, avoiding
you the burden of the answer. Sorry for that. My first use of these
files is to keep the matching between our remote EC2 machines and our
local Eucalyptus ones, trying to make sure they are based on exactly
the same base images, and this file is just perfect for that, a plain
'grep' doing the trick (it is actually in that context I detected the
bug mentioned above).
Yes, definitely my argument was referring to better keep N very small
files + one small file than querying (N+1) very small files each time.
By the way, not that it impacted me, but I noticed some
"irregularities" in the naming of beta and alpha releases. For
example, for Lucid you used "alpha-1", "beta-1" in the path an "beta1"
in the filename. Then, still in Lucid, you provided another path,
"beta1". Now go to Maverick and you again have "alpha-1" path with no
"alpha1" name. Is that supposed to be the logic from now on? I mean a
'SUITE-server-uec-ARCH.tar.gz' generic name, whatever the release,
specifics being given by the path?
Thanks again for all your work Scott. It definitely made our life
easier. Ubuntu rocks!
On Aug 30, 5:29 pm, Scott Moser <smo...@ubuntu.com> wrote:
> Well, I was just trying to avoid duplication of data, and aiding people to
> get really only what they need/want rather than everything. Its not
> perfect, and the overhead of multiple requests is probably worse than one
> request for more data since we're dealing with pretty small files.
> The usage that I was designing for was basically people pollinghttp://uec-images.ubuntu.com/query/released.latest.txt, and determining
> if they were out of date based on that data. If so, they would do further
> queries to figure out what actions needed to be done.
> That said:
> http://uec-images.ubuntu.com/query/released.latest.txt > gives you a row for each unique build in the released 'stream'.
> The 'released' stream contains builds labeled 'alpha-*', 'beta-*', and
> 'rc', and 'release'. The other stream is the 'daily' stream.
> The primary key for a build is a combination of the 4 fields:
> suite, build_name, label, serial.
> ie:
> hardy server release 20100827
> That file has 7 fields
> The first 4 of the key (suite, build_name, label, serial)
> Followed by arch, path, and 'suggested name'.
> So, by walking each row in the released.latest.txt, and getting the
> appropriate released-dl.txt, you can get the data you want.
> Ie, this will get you what you wanted:
> burl="http://uec-images.ubuntu.com/query"; str="released"
> wget -q "${burl}/${str}.latest.txt" -O - |
> while read suite bn lab ser; do
> wget -q -O - "${burl}/${suite}/${bn}/${str}-dl.txt" ;
> done
> Alternatively, the data on uec-images is available via rsync. So, the
> following will also get you want you want:
> rsync -a uec-images.ubuntu.com::uec-images/query .
> cat query/*/*/released-dl.txt
> By the way, not that it impacted me, but I noticed some > "irregularities" in the naming of beta and alpha releases. For > example, for Lucid you used "alpha-1", "beta-1" in the path an "beta1" > in the filename. Then, still in Lucid, you provided another path, > "beta1". Now go to Maverick and you again have "alpha-1" path with no > "alpha1" name. Is that supposed to be the logic from now on? I mean a
The practice of '-' in the directory name, but not in the filename comes from the other code that I use in the publishing of images. Ie, that is (or at least should be) consistent with other ubuntu releases.
It appears to me that the query data consistently references files using 'beta-1' rather than 'beta1'. One way or another, the query data gives you a hassle free path into all of this. You shouldn't have to worry about a 'label' -> 'filename' or label -> path conversion as the data provides it for you.
In query, there will not be hyphens in label names. (but I won't make promises about path names).
> 'SUITE-server-uec-ARCH.tar.gz' generic name, whatever the release, > specifics being given by the path?
I did change the file naming convention once, and that ticked some people off. I believe that we should be able to stick to <suite>-<buildname>-uec-<arch>.tar.gz for a while. Again, hopefully even if I changed it, the query would provide you with the correct data.
One thing that I'm not terribly happy with is that for dailies, the <suite>-<buildname>-uec-<arch>.tar.gz contains files named like "<suite>-<buildname>-uec-<arch>", but for published builds, the contents of the tarball are still '<suite>-<buildname>-uec-<arch>', but the tarfile name has changed.
That is a side effect of the fact that "promote-daily" doesn't change anything that was MD5SUMed. Ie, the build that was a daily produced a <suite>-<buildname>-uec-<arch>.tar.gz , and that was simply copied and renamed to ubuntu-<suite>-<version>-<label>-<arch>-<buildname>-<serial>.
The inconsistencies that you find are a result of happening as they went, not purely designed from the beginning.
So, in summary, if you use data in /query, then you shouldn't have to worry about this stuff.