GitLab

What I did was create node table entries for the three SPP nodes.
These are designated as local, shared nodes, reserved to a holding
experiment. This allowed me to use all of the existing shared node
pool support, albeit with a couple of tweaks in libvtop that I will
not bother to mention since they are hideous (another thing I need to
fix).

The virtual nodes that are created on the spp nodes are figments; they
will never be setup, booted or torn down. They exist simply as place
holders in the DB, in order hold the reserved bandwidth on the network
interfaces. In other words, you can create as many of these imaginary
spp nodes (in different slices if you like) as there are interfaces on
the spp node. Or you can create a single spp imaginary node with all
of the interfaces. You get the idea; its the reserved bandwidth that
drives the allocation.

There are also some minor spp specific changes in vnode_setup.in to
avoid trying to generalize things. I will return to this later as
needed.

See this wiki page for info and sample rspecs:

https://www.protogeni.net/trac/protogeni/wiki/SPPNodes

its a delay node.

the tb-set-node-os command with a second optional argument; if that is
present, the first arg is the child OS and the second is the parent OS.
We add some new features in ptopgen (OS-parentOSname-childOSname) based
off a new table that maps which child OSes can run on which parents, and
the right desires get added to match.  We setup the reloads in os_setup
along with the parents.  Also needed a new opmode, RELOAD-PCVM, to handle
all this.

For now, users only have to specify that their images can run on pcvms, a
special hack for which type the images can run on.  This makes sense in
general since there is no point conditionalizing childOS loading on
hardware type at the moment, but rather on parentOS.  Hopefully this stuff
wiill mostly work on shared nodes too, although we'll have to be more
aggressive on the client side garbage collecting old frisbee'd images for
long-lived shared hosts.

I only made these changes in libvtop, so assign_wrapper folks are left in
the dark.

Currently, the client side supports frisbee.  Only in openvz for now, and
this probably breaks libvnode_xen.pm.  Also in here are some openvz
improvements, like ability to sniff out which network is the public
control net, and which is the fake virtual control net.

virtual topology, get tickets, redeem tickets, map onto Emulab nodes.
Work in progress!

This broke control net firewalls.

Make my best guess attempt at what the code should do; i.e., there are still
cases where we probably shouldn't allow such command lines.

like shared host nodes. These are removed when the experiment swapped
or modified, just like the rest of the virtual state.

the pport is null; a trivial link. Need to revisit this later.

Also a shared node change; do not update remaining_bandwidth when
doing a swapmod of the shared_node experiment, if the node was already
in the experiment. Might have existing vnodes using those interfaces.

existing Lans of the same name.

shared host.

Remove some code that snuck into last revision; does not work yet.

Add a solution option that is similar to impotent mode,
but spits out an xml file with the solution. Used by the CM.
A couple of little fixes.

(we used to return one just for veth encap).

allocated.

cluster nodes. Not going to try and describe all these changes.

Note that I have not back ported this into the old assign wrapper. We
move inexorably forward.

Worth mentioning:

* Users get VMs only on shared hosts.

* Multiple experiments from multiple projects can share a node.

* Nodes that are acting as shared hosts are in a holding experiment
  and have a tag in the reserved table. All of the links in the
  experiment are tied together in one giant super vlan. We then
  multipleax over that using our standard mechanisms (veths, vlans,
  whatever).

* Lots of complication in the link setup code for dealing with links
  between a virtual node on a shared node, and a private physical
  node. Requires additional vlans and trunking between those
  interfaces. To make life easier, all of the links in the afore
  mentioned super vlan are trunked in dual mode.

* I had to change Mike's code that does the encap determination, since
  openvz nodes cannot do veth encap. I add some new osfeatures
  (veth-ne, veth-en, and vlans).

* On and on and on ...

use attributes.

that depending on different versions of assign to find the exact same
solution on the same top/ptop, is asking for trouble. And in fact, the
new assign is slightly different and the solutions do not match.

So the idea we came up with is to first run the old version, and then
fixnode the results for the input to the new version. assign should
run cleanly and the results will be the same. Then reverse the
situation and run the new version, then fixnode those results for a
run of the old version. Then compare all the results.

we no longer call os_select as a shell script but rather as a library
routine, which save bundles of time on big experiments.

physical state compares properly. Also some actual fixes.

experiment in regression mode, and then compare the results.

wrapper/mapper, so that DB tables are ordered in a consistent manner,
and can be easily compared with just diff.

Also some actual fixes to the mapper, as a result of regression mode
comparisons with the wrapper.

mapper wrapper. In regression mode, the wrapper/mapper proceeds
normally, creating a .vtop file, and then running assign with a fixed
seed. If the wrapper and the mapper agree on the .vtop file, then the
solution from assign should be identical.

The wrapper/mapper then proceeds normally, reserving resources and
making all the DB changes. Needless to say, this has to be on a
private copy of the database, with all nodes free. Creating that DB
was a tale in its own right.

At completion, call the existing BackupPhysicalState() function that
we use in swapmod, and write all the physical tables we have changed
(just the rows corresponding to the experiment of course). The delete
all that state, and free the nodes.

If everything is working correctly, those physical tables should be
identical when created by the mapper or the wrapper.

Of course, its not quite there yet. I have a few things to fix up
before diff -r produces no results.

Oh wait, the line moved, I see Kevin has made changes to
assign_wrapper that I need to pick up.

Linkdelays also done.