(including the MEs and FEs and geni racks and ...):

* Move the also-notify statement from top level into the external view.
  I noticed when I had debugging on the Mothership, that we are getting
  gobs of notifies for the 172 zones (16 of them) from all over the
  place over and over, which the Mothership just ignores. This was bad
  for two reasons; 1) it was exploding the debug file, and 2) it was
  tripping the "quota" for the actual zones that need to be transferred,
  so those were deferred for several minutes. This was especially
  annoying while I was debugging cause the zone I cared about was always
  at the end and deferred. Cue the thumb twiddling.

  OK, but this did not work, all those notifies kept getting sent. Turns
  out we also need a "notify explicit" at top level to turn of automatic
  notify of all zones to the masters.

* For the ME/FEs I added a super hack to make sure that the special
  internal 10.10.10.0/29 network is served by the internal view. I need
  to generalize this at some point.

* Add a commented out logging section so I do not have rediscover the
  syntax next time I have to drop into this pit of hell.

And with all of that, the original problem that sent me down this road
was still not solved. That is another tale that needs to be told, stay
tuned.

Kernels and initramfses are getting larger and larger, and because
pygrub uses a tmpfile inside /run, and because dom0_mem is usually
quite low (and thus so is the default /run tmpfs allocation), we
have to increase the max size of /run.  128M should be fine for
now, and that is below dom0_mem in all of our cases.  Someday we
might need to be more intelligent about how we set /run's size.
(Right now this is only necessary for UBUNTU20-64-STD on a pc3000,
I think, but I happened to hit it because my grub.cfg reverted to
dom0_mem=1024M after a package install.)

it no longer builds.

branch, even if the user had switched branches.

The named changed between metis4 and metis5.

rule for news.php.

ago, so now the ssl certificate for apache will actually be correct
to start with.

we need it running for the proxied console support.

Metis4 is gone, replaced by Metis5.

reservations, check first to make sure they can all be approved.

It is no longer fixing up FreeBSD partitions because the check_disklabel
script doesn't seem to work anymore. It seems to be having problems with
the FreeBSD disklabel. So if it fails, just assume the DOS partition is
the root partition. That works in our current layout (i.e, the last 10
years or so).

Now if I can just convince Dave to remake the Ubuntu images...

Didn't really give it much thought when I said that. Anyway, the browser
consoles on the buses are tricky cause ops does not have a routable IP,
and that is where shellinabox runs, using tiptunnel to connect to the
capture on boss.

I did not want to move shellinabox to boss, its a big pile of gross code
that I treat as a black box.

So ... I found my thimble full of mod_proxy and mod_rewrite skills,
and managed to use apache on boss, to proxy over to the shellinaboxd
on ops, which talks back to boss. Will this circle stay unbroken?