GitLab

mark the images to run on them. Not a great way to do this; needs more
thought.

entries.

node.

(mngmnt).

localize_mfs, the later so we can dhcpd to listen on other interfaces.

"You people and your frickin relative source dir paths."

Note that this only makes sense for singlenet==0 where we can control the
speed of the inner elab control network.

Figures out control net interface, sets up a new bridge, and moves
the control net IP to the bridge. Also updates the controlif file
to point to the new bridge.

Known limitation: Does not preserve IP aliases from control net
interface

the os subdir into the clientside/os dir.

Mike, use the --follow argument to git log to traverse across the mv.

Should be harmless (i.e., not change any rows) but I want to get it out
of here to avoid confusion.

This call is already done in script1.sh. My intention in putting it in
rc.frisbee was to phase out script[12].sh. But there is enough magic in
those scripts that I am just going to leave them be. In particular is the
constant restarting of tcsd to avoid hangs.

This way we can be 100% certain we are running the binary we think we are.

emulab-install, or from the current root password.

install script, but I pulled out to create an independent script).
This works on both freebsd and linux based MFSs. The intent is to do
all of the localization automcatically for site admins, so that they
can import new MFSs more easily. This is also used from the new
install code to bring in the initial MFSs and localize them.

Here is what we localize:

* The timezone is copied from boss:/etc/localtime to mfs:/etc. Ryan
  says the upcoming version of the linux MFS will actually use
  localtime. 

* Copy boss:/usr/testbed/etc/{emulab.pem,client.pem} to mfs:/etc/emulab. 
  The former is for TPM, the later for the ssl version of tmcc.

* Copy out boss root ssh keys (pub) to mfs:/root/.ssh/authorized_keys.
  In an ElabInElab we take care to combine with outer boss keys.

* Copy out the image ssh host keys. These are the keys that we put on
  every image to avoid the ssh host key change sillyness. See notes
  below on how these keys are initialized on an existing emulab. The
  keys are copied from boss:/usr/testbed/etc/image_hostkeys to
  mfs:/etc/ssh directory.
  
* Initialize the root and toor passwords from a new sitevar named
  images/root_password (which is the encryption hash, not plain
  text). See notes below on how this sitevar is initialized on an
  existing emulab.

About initializing the host keys and the root password hash ... I
added a new update script (27) that will go out to the current frisbee
MFS and mount it, grab the current keys and password hash, and put
them into place on boss. At the moment I only look for a FreeBSD
frisbee MFS, since not too many people are running the linux mfs, and
this was hard enough as it is!

For a new installation, a new install phase script will build the them
and install into /usr/testbed/etc/image_hostkeys. I have not dealt
with the password yet.