GitLab

hardwire to apache22).

David is going to add this to prepare so we do not create images with
stale repos on them.

Mainly, this means tag create messages now have tag URLs; and that
annotated tag create messages now include more tag metadata, as well as
the referenced commit.

jacksmod is now a dependency of Jacks. In order to prevent loading problems, Jacks itself no longer loads jacksmod. So it is up to the surrounding page to load it properly now.

execute services. So for tainted images use the older method.

EXPORT_PAT for different mount output.

have to wait, they are instantly mounted on boss, but do not show up in
mount output.

Don't know why we did not hit this before. Maybe the node_usage DB
got into an odd state recently where there were some extra records?
Who knows. I compared the .dat files produced from old and new and
all the historic records that this script produces are the same.

on boss cause that is where the actual file systems are.

link to user. This is a different then the Forgot Password link, in that
it does not require half the key to be in the browser cookies.

into a whacked out state cause of earlier errors. Be careful.

the file systems are back over on boss, and chflags does not work over
NFS. I think we need this for the powder nodes, so I will turn it off
for now, but come back later to make it happen.

list of nodes that can be imaged.

* Makefile changes to build and install nossl versions of capture and
  console on a rack control node (or more generally, a physical node
  hosting boss/ops VMs that are not built on our XEN49 image).

* Add -I (insecure) option to capture, that listens on localhost only.

* Add systemd startup files for capture on ops and boss, I tested these
  on Ubuntu18.

Basic instructions:

* Clone the emulab-devel repo to the control node.

  git clone https://gitlab.flux.utah.edu/emulab/emulab-devel.git

* On the control node, install the libssl devel code:

  sudo apt-get update
  sudo apt-get install libssl-dev

* configure and build capture. Note that the obj-clientside directory might
  already exist, you can just rm -rf the directory.

  control> cd ~elabman
  control> mkdir obj-clientside
  control> cd obj-clientside
  control> /path/to/emulab-devel/clientside/configure
  control> make rack-control
  control> sudo make rack-control-install
  control> (cd os/capture; sudo make rack-control-startup-install)

* start capture.

  control> sudo systemctl daemon-reload
  control> sudo systemctl start capture-boss
  control> sudo systemctl start capture-ops

This has been bothering me for a while; A single execute service
operates like a traditional Emulab startup command, with output going to
/proj/pid/exp/eid/logs, which is a pain in the ass cause I don't know
the name of the file and there 1000s of file in the directory.

But when there are multiple execute services, we wrap them in a script
and redirect the output to easy an easy to find spot; /var/tmp. Much
nicer.

Now we always wrap them up in a script so the output files go to
/var/tmp. And drop a note in the original file that says where to go
look.

The script is written to /proj/pid/exp/eid, but thinking to the future
when there are no NFS mounts at all, we now bundle that script into a
little tarball and append that to the install services. Tarballs already
handle a no-NFS world, asking the web server for the file. QED

* Stop using the ALWAYSUP state machine for switches, this causes ISUP
  to always get sent, which in certain cases, results in stated
  rebooting the switch!

  Added new ONIE state machine, which handles the way switches actually
  boot into ONIE first and then does the bootinfo/grub dance, or does a
  reload or does admin mode.

* Do not send PXEBOOTING from ONIE; this was a mistake, it throws us
  into the PXEKERNEL state machine, which sometimes results is stated
  rebooting the switch!

  We still use PXEWAIT (it is sent by bootinfod), since that is the
  "waiting" state that is wired into a lot of Emulab, it just happens to
  now be a state in the ONIE state machine, so its legal.

* Fix a bug in libossetup, that was fooling libossetup_switch into
  thinking the wrong thing.

* Add some timeouts to the libosload_mlnx code, sshd sometime refuses to
  answer after a failed login. Strange.

* Fix a fork() problem in the switch reload code; gotta call exit, not
  return! This was wreaking subtle (okay not so subtle) havoc in
  libossetup.