GitLab

How simple you all seemed compared to the terrible links ahead...

Basic framework is in place. Availability and printability is codified into
functions now. Node.pm modified so that lookups can be done in bulk for
efficiency. When making XML, we now create XML trees rather than printing
out XML fragments.

Send email when a prereserve is fulfilled.

idea).

New script and table to manage node pre reservations. Lets just look
at the script.

To create a reservation:

    myboss> wap prereserve -t pc850 testbed 2
    Node reservation request for 2 nodes has been created.

To see the reservation status for testbed

    myboss> wap prereserve -i testbed
    Project         Cnt (Cur)  Creator    When               Pri Types
    -------------------------------------------------------------
    testbed         1 (1)      stoller    2011-08-12 12:39:07 0   pc850

    which says 1 node is pending and 1 node has already been
    pre-reserved. 

To clear the above reservation request (and optionally, clean
reserved_pid from the nodes table).

    myboss> wap prereserve -c -r testbed

    The -r is optional, otherwise just the reservation request is
    cleared, and nodes continue to be pre-reserved to the project.

To see a list of all reservation requests:

    myboss> wap prereserve -l


So, when a node is released in nfree, we look at the reservation
status for the node and any pending reservation requests.

1. If the node has a reserved_pid and that request is still pending
   (still in the table), nothing is changed.

2. If the node has a reserved_pid, but the request has been cleared
   from the pending table, then clear reserved_pid.

3. If reserved_pid is null, and there are pending requests, then pick
   the highest priority, most recent dated, request, and set
   reserved_pid to that project.

Options:

* -n <pri> - is how you set a priority. Lowest is zero, choose a
  higher number if you want this reservation request to be considered
  before others. In a tie, look at the date of creation, and use the
  oldest.

* -t <typelist> - a comma separated list of types you want to
  consider. Types are considered in order, but not in the fancy way
  you might imagine.

1. Support for "one-shot" PXE booting ala the one-shot osid. Switches to
   pxelinux to boot WinPE and then switch back after done. Painful now
   because we have to HUP dhcpd everytime we change the PXE path, but we
   may be able to fix this in the future by going all-pxelinux-all-the-time.

2. Added pxe_select, analogous to os_select, for changing the pxe_boot_path
   including the one time path.

3. Added the WIMRELOAD state machine to shepherd a node through the process.
   Still has some rough edges and may need refining.

CreateVnodes().

We set the passed in ref to "" in the case of no log (rather than leaving
it undefined).

URNs in component_ids should now always refer to the physical node rather than the virtual container.

This allows other things that don't need to change the OS, but *do* need
to make sure that a mode transition gets setup if necessary, to setup that
transition too.  Currently the only case that needs this is push-based
reloads, when the reload has been setup (and then not performed because it
was optimized away).

This is completely reactive and based on no reasoning. A common case of the
somebodys-been-messin-with-my-memory bug, is that we start seeing node_ids
with null chars at the end. Here we will send a stack trace to testbed-ops
if this happens (and we strip off the null). I am hoping to narrow down
where this happens.

Lbs

This is a fer-now hack.  Perl 5.10 has issues with bogus taint check
triggers that appear quite often when using the SelfLoader.  Now if you put:
    SELFLOADER_DATA=""
In your defs-* file, it will disable the uses that cause problems.  Yes,
the configure script should figure out if this is needed for you, but...
another time.

as Node->CreateVnodes() instead of Node::CreateVnodes().

if its already set up, as is the case when coming from wanodecreate.

starting any one of our scripts can take a second or two. That time is
spent including and compiling 10000s of thousands of lines of perl
code, both from our libraries and from the perl libraries.

Mostly this is just a maintenance thing; we just never thought about
it much and we have a lot more code these days.

So I have done two things.

1) I have used SelfLoader() on some of our biggest perl modules.
   SelfLoader delays compilation until code is used. This is not as
   good as AutoLoader() though, and so I did it with just a few 
   modules (the biggest ones).

2) Mostly I reorganized things:

  a) Split libdb into an EmulabConstants module and all the rest of
     the code, which is slowly getting phased out.

  b) Move little things around to avoid including libdb or Experiment
     (the biggest files).

  c) Change "use foo" in many places to a "require foo" in the
     function that actually uses that module. This was really a big
     win cause we have dozens of cases where we would include a
     module, but use it in only one place and typically not all.

Most things are now starting up in 1/3 the time. I am hoping this will
help to reduce the load spiking we see on boss, and also help with the
upcoming Geni tutorial (which kill boss last time).

attempts by protogeni code.

commit.

use when (re)loading a node.

(class is switch or role is testswitch). Also add a node_type
attribute for all existing switch types: forwarding_protocols=ethernet

partition table lookups here. Used by os_setup, etc.

severely flaky.

reserved yet, and make it reserved into the provided experiment.

move them into hwcheckup.

Make sure nodes are imageable, otherwise straight to hwdown.

Minor fix to Node::MoveReservation().