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

Fix: subnode_of tags now contain the full URN in RSpec advertisements rather than just the node name.

Added an optional subnode_of option to both advertisements and requests. Ptopgen now generates the subnode_of relationship in Geni mode.

Change location sitevars to general/default_* rather than protogeni/default_* as these may be more widely applicable than just Protogeni code.

Changed flag output to attributes in ptop case to conform with Tarun's changes from a few months ago.

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.

If in geni mode, do not print links from widearea nodes to the 'internet'. These will be printed based on interfaceroles anyhow.

ptopgen now emits 'role="experimental"' on interfaces. rspec-ad.rnc will now accept such advertisements.

The 'internet' node is defined and has a single interface called 'border' and is marked by the '<cloud />' element. All control interfaces with publicly routable IP addresses (not in one of the three private IP spaces) have a link to the 'internet'. Those interfaces now also contain a 'public_ipv4' attribute which contains their address. The modified schema validates this.

ptopgen now emits 'role="control"' on each control interface definition. The rspec now validates against it.

Added a '-1' option which can be used to get the rspec of a single node. Note that this doesn't attempt to optimize away any database lookups.

have that information in the widearea_nodeinfo table.

invalid XML.  We previously used a workaround in GeniSES.pm to keep
the demo alive.

find out how much bandwidth is currently reserved, so that
we can modify the bw on the physical interfaces appropriately, for
shared-nodes.

genimode.

I also wrote code to use the bandwidth slot of the vinterfaces table,
but moments after I wrote it I realized that the vinterfaces table has
benn cleaned by the mapper (RemovePhysicalState()) by this point, so
that will not work.

I do not have a viable solution yet.

shared. This cuts down on the rspec size for the demo; huge
otherwise.

Another temp hack; strip out interfaces that are not expt/ctrl to
prevent clutter and breakage in the interface list.

Added a bogus "(null)" interface to all switches to mollify assign. Still need to really solve this problem in a backwards compatible way.

from the main site; only nodes with reserved_pid set.

	if ($sharing_mode{$name}) {
	    print "  <exclusive>false</exclusive>\n";
	}
	else {
	    print "  <exclusive>true</exclusive>\n";
	}

packing to favor nodes that already have vnodes on them. This can also
come from the node_attributes table.

Then I disabled these changes cause it was causing me problems; the
shared nodes were not getting picked. Needs more thought.

CM URNs for links are now shown rather than UUIDs. Fixed bug in the 'available' tag which showed all local nodes as unavailable in rspecs. Fixed selection criteria so that all nodes are shown in an rspec rather than just the available ones.

Fixed to accomodate Tarun's tweaks to the schema which changed a few thing from elements to attributes and used interface_ref.

ptopgen now produces output which conforms to the rspec advertisement schema. It produces URNs rather than UUIDs.

* New option -h to turn on shared nodes; set by the mapper wrapper
  when it sees that an experiment wants to use shared nodes.

* When using shared nodes, look for physical nodes that are reserved,
  but have their sharing mode set. The node must also be up.

* Shared hosts nodes get the feature pcshared:1.0 in the ptop
  file. libvtop adds the corresponding desire.

* On the shared hosts, do not spit all the OSIDs, but just the one it
  is actually running. This allows the user to indicate what kind of
  shared node they want via the osid. libvtop adds that as a desire.
  Crude but effective.

* Use the interface_state table to get the remaining bandwidth on the
  interfaces for shared hosts, so that assign does not oversubscribe
  links.

* Some initial code to look at the load averages on shared nodes, to
  guide the packing. Not actually operation yet.

  At the moment packing is still controlled by the collocate factor
  from the experiment (NS file). Eventually we want to use the load
  averages above, or other feature/desire mechanisms to control
  packing.

<available> is now based on reality. interface node_uuids are also real. TODO: link_uuid, virtualization_type, (null) interface name.

ptopgen now generates a file which validates under the new rspec schema. ptop/ptopxml are identical. Some of the rspec output is still placeholder.

Special case for interconnects to make interface names source and dest rather than source:source/source_if and dest:dest_if

Finally checked in a refactoring of ptopgen's output code. This includes making rspecs mor rspec-y rather than just a slight variation of the ptop xml format.

node_attribute with key='protogeni_exclude'. Primitive, but since
there is a web interface and a backend script, its easy for remote
sites to use.