On 2016-02-03 1:37, Joe Brockmeier wrote:
On 02/03/2016 09:24 AM, Hannes Schmidt wrote:
So EOL implies that no one should ever be able to launch a VM for it?
Makes no sense to me.
It's no longer receiving updates, so that does actually make sense.
>
Can you describe the use case for an EOL release that might persuade us
that we should continue paying for the storage required to host EOL AMIs
and risking that users might deploy them without realizing they are no
longer receiving security updates?
The expectation in AWS is that released images stay up forever. If
you're looking for an example of why that is useful, I used to run
compute jobs with no access to the Internet at all; I literally could
not log into the instances even if I wanted to. Reproducibility of
results was more important than updates, vulnerabilities or not, even
while the releases they ran on were still supported. I could point
everyone who wanted to run those workloads themselves to a well-known image.
That aside, let's be realistic here: storing 3GB of snapshot data in
all ten EC2 regions costs $3.04 per month. Even after accounting for
two architectures and the past few releases' PV-vs-HVM split, I could
host every Fedora Cloud release ever produced for less than my personal
AWS bill. If cutting costs is that important, are the final release
images really the place to focus? It is clear that some people find
them valuable.
--
Garrett Holmstrom
_______________________________________________
cloud mailing list
cloud@xxxxxxxxxxxxxxxxxxxxxxx
http://lists.fedoraproject.org/admin/lists/cloud@xxxxxxxxxxxxxxxxxxxxxxx
