A bit of a short week this week, as thursday is 4th of july holiday in the US and I am taking friday off, but still a bunch of goings on.

koji was having some bad days on sunday/monday. Turns out that our block_retired script that runs at the start of every rawhide compose figured out that all the packages in epel7 were now end of life/retired and so it started trying to untag and block all of them. Unfortunately this script is normally meant for a few packages being retired and it wasn't really setup well to handle trying to deal with 15k packages at once. Also, we don't actually want it to untag those packages, we want to keep them around for historical reasons mostly. So, I managed to figure out what was going on and stop it, but it had untagged about 10k packages by then. Will look at cleaning this up more next week.

There was some discussion around the new ssh host key on fedorapeople.org. I had failed to announce that it had changed and a few folks (very rightly!) asked if the change was expected. I actually thought about just preserving the old host keys, but they were made 10+ years ago now, so I figured it would be time to generate new ones (and prefer the newer algo also). I did then send an announcement to devel-announce about it. Out of this docs were improved and I tried to push the idea of trusting the fedora infrastructure ssh CA. This just requires you to add the CA to your .ssh/known_hosts and it will trust host keys that are signed by it. This includes all fedora infra hosts. You can of course also use SSHFP, but that requires some ssh settings and confirming that you are using a dnssec enabled resolver, so it's a good deal more work. Anyhow, hopefully this host key will last us at least 10 years too.

There was a bunch more rhel7 vm cleanup that happened. We are still sadly not 100% done on that, but there's only a few left, almost all just internal things and blocked by various things that will hopefully unblock in the next few weeks and we can get them all dealt with too. I think the rhel7 EOL has been harder on us because of the python2->python3 move. Not that I disagree with it, but it's just been a lot of development work to move things up and get them back in a maintained state. Really it's our own fault, but it's hard to prioritize things when they are working fine.

Finally, just a note about next week: I'm off monday, not doing anything fun, but going in to have a molar pulled. So, I am likely to be a bit grumpy later in the week (I can't even have coffee for a while, thats going to hurt).