GitLab

This is pretty minimal "support", but it is working.  I have some
uncommitted fixes for event/linktest/iperf, however, as mentioned in
issue #351.

The organization of the code follows the guidelines set by the
ubuntu version extremely closely. Big differences to notice are
some of the package name changes switching from apt-get to apk,
the requirement of the --disable-ssl flag when running the configure
command for pubsub and the methodology required to build a custom
apk package to use our custom runit. Apk has lots of issues and
argues against building a new .apk package as root, so
runit-artifacts.sh creates a new user, gives them super user permissions,
adds them to the abuild group and runs the runit-packager.sh script
as this new user. Design of this solution was focused on doing as
much as possible while remaining in root. Enjoy!

Replaced by 40-emulab.

With it, iSCSI blockstores on Ubuntu16 and CentOS would not work.

Also, add a print statement in the unmount code that Hussam added so
that we can see how often this (still?) happens.

Many LV segments fail on "dmsetup remove" because they are still mounted. Need to unmount them before the remove operation.

Make sure the FS create actually fails before you cleanup and remove
the new volume...

We want it processed before the default file so that we can stop
Emulab (local5) messages from going into "syslog".

We were setting up ipod and then calling "tmcc fullconfig" which re-fetched
the ipod info causing a reset of the one-time secret. Thus the kernel had
the wrong secret.

I'm not sure if all containers will want stdin opened or attached on
console (i.e. for the master process), but for now we'll just turn it
on.

This means that users can still use our ssh urls to reach containers
that don't run sshd.  We run a private sshd that has Ports to listen
on, and Match blocks containing ForceCommand directives, which run
docker exec $vnode_id <shell>.  User can configure which shell.

Turns out the Linux raid tool (mdadm) will pick up fragmentary RAID
metadata and complain mightily.

We do this similarly to Xen.  There's a new script (container2pty.py)
that attaches to the Docker container, via the docker daemon, and
exports its stdio as a pty.  Then we run capture on a symlink to that
pty.  New options to capture tell it to keep retrying to open the pty
maxretries times (we invoke with infinitely many retries); and to not
prepend /dev to the device string.

Not sure why I singled out grub1 to not add that argument, it is
documented to work and does. By not setting it, old CentOS images
that used grub1 would hang at boot time for > 5 minutes causing a
stated timeout and reboot. Once stated had tried that 3 times, it
would stop doing the reboot and eventually the node would come up.

I have no idea why it was hanging unless the console output was
going to the real serial port (com1) and that was either generating
garbage input causing grub to go interactive or causing output to
block.

This was the case where we would make a copy of the dom0 root FS
to create the vnode's FS.

(This might be a bandaid for another problem with docker's attach;
I seem to get some garbage from some container stdout/stderr, especially
in this case.  Probably encoding that I'm not handling yet; but this
papers over that.)

(Of course, there is no standard perl Docker Engine API, so I wrote my
own thin perl library, basically error handling and JSON marshalling
goo.  But had to deal with LWP/http over UNIX socket to Docker, which
is whacky; fortunately, there was a module, but I can confirm that LWP
is a pain when you get off the beaten track.  Anyway, I added a little
client that is useful for debugging.)

Also, I refactored things a bit so there is now an emulabize-image
script that can be run to test emulabization of docker images
separately from vnode creation.

Also, add a specific Docker version of the prepare script.  This
actually gets run in image commit as an ONBUILD instruction.

Also, always copy the image's master passwd files into /etc/emulab
instead of using those installed by client-install.  Those just don't
apply at all here since we have no idea what the base image is; the
uids/gids in the clientside dirs could be a complete mismatch for the
image.

node:card.port format, not node:card. Be that as it may, snmpit needs to
change at some point to print node:iface instead.

Just don't build it for now. Will have to fix iperf at some point.

No more hacky python proxy to the old django interface! No longer needed,
as the REST API now covers everything we do. Even if we wanted to keep
around the old API stub, it would need work as FreeNAS has switched to
python 3 and many of the old API python modules are gone or renamed.

If one blockstore name was a subset of another (e.g., "foo" vs. "foo2")
then the check code could get confused and not mount some iSCSI blockstores.

We were not returning the correct number of values.