Post Move Stuff
Mike McGrath
mmcgrath at redhat.com
Wed Dec 16 04:08:01 UTC 2009
So the move itself is over but there is still a lot of work to be done.
At the moment some of our normally redundant services (koji, vpn) aren't
redundant. Also I'm pretty sure puppet is still failing on some hosts but
that we can fix at our own pace.
Also just a note, smooge and I are likely going to be in recovery mode the
next couple of days. So if something seems broken please open a ticket.
So what is left?
- We started renaming everything, we'll need to finish that (involves
renames, re-keying things, etc)
- New network map. We went from essentially having a single network to
having 3 networks:
- A build network
- A storage network
- A public network
- Training on one of our new server types, bxen*. These hosts are
dedicated to build and releng activities. This was done for a couple
of reasons, most of which are organizational. It will also allow us to
more easily predict growth needs for the buildsystem in the future.
- Figure out what to do about proxy servers in PHX2. We can go a load
balancer route or a heartbeat route but I'm not totally convinced we
need two proxy servers in PHX2 like we had in PHX1 though, because of
the way network routing still works we'll have to figure out something
HA
- QA - The new QA boxes are ready to be configured, I'll be working with
jlaska on this. It's the first kind of hosted by Infrastructure but
not really run by infrastructure set of boxes. Similar to how the
releng boxes work but the QA team is less close to Infrastructure then
release engineering is. This will involve training and some new
policies.
- Host certification - this is something I've been working on but not
enacted yet. Mostly a solid lookover everything based on a recent CSI
doc.
http://infrastructure.fedoraproject.org/csi/host-lifecycle-policy/en-US/html-single/#HostLifecycle-Host-Recertification
In some organizations this the certification process will help bring
about accountability. For us it's more about knowing what's going on,
it's not like if someone accidentally certifies a box wrong they'll be
in trouble but in our case a second pair of eyes will help. Even in
the first trial run I did with smooge he discovered something I missed.
There's also a lot of little things to do, especially with verifying
things like IPTables and monitoring.
-Mike
More information about the Fedora-infrastructure-list
mailing list