On 05/15/2017 08:59 AM, Scott McCarty wrote:
On 05/15/2017 08:28 AM, Dusty Mabe wrote:
I would encourage you to make this image "timeless". For the lab we did at
summit someone could theoretically download the qcow2 years from now and still
run through the entire lab. We baked all of the container images we needed as
well as all rpms we needed for 'builds' into the image and also tested it
completely offline.
Yeah, I really, really like this idea. I need to figure out how to get
the RPMs baked in. I am quickly realizing that I need to have a Red Hat
downstream version of this lab and an upstream version because the RPMs
for the rhel-tools image aren't redistribute-able :-( without breaking
the RHEL legal rules.
What did you use to cache the RPMs? Some kind of proxy?
I initially set up a squid proxy and ran through the entire lab with the
container builds pointing to the squid proxy. Then I just used the logs
from the squid proxy to grab the NVRA of the rpms we needed and then used
yumdownloader (or dnf download in your case) to get them.
I wouldn't bother trying to pull the actual rpms out of the squid proxy
cache. It's in some format that is optimized for fast access and not trivial
to get the actual files back out of it.