GitLab

Keeping them up to date throughout the node lifecycle is not a lot of fun...

Also quiet down the EmulabFeatures library where we are using it
to detect external management.  Too many "checking for feature ..."
messages.

In support of the collaboration effort with CHPC, I've added a check
and actions for a new feature, ExternalNodeManagment.  When swapping in,
the swapper, group, and experiment are checked for the feature.  Since
there doesn't appear to be a way to ask for an experiment feature in the
parser, I'm punting for now and just letting the experiment inherit this
feature from the user and/or project/group.

The effect of this feature is that nodes will be omitted from the
dhcpd.conf file when allocated to a marked experiment, and put back
in when freed from that experiment.

We will also use this feature to force nodes to reboot even when they
are in PXEWAIT during the OS setup phase.  Ultimately the point is for
something else to take over the boot process and manage the experiment's
nodes for the duration of the experiment.

we update last_ext_act, since this is how idle determines staleness.
Not sure this is the best approach though.

This commit is intended to makes the license status of Emulab and
ProtoGENI source files more clear.  It replaces license symbols like
"EMULAB-COPYRIGHT" and "GENIPUBLIC-COPYRIGHT" with {{{ }}}-delimited
blocks that contain actual license statements.

This change was driven by the fact that today, most people acquire and
track Emulab and ProtoGENI sources via git.

Before the Emulab source code was kept in git, the Flux Research Group
at the University of Utah would roll distributions by making tar
files.  As part of that process, the Flux Group would replace the
license symbols in the source files with actual license statements.

When the Flux Group moved to git, people outside of the group started
to see the source files with the "unexpanded" symbols.  This meant
that people acquired source files without actual license statements in
them.  All the relevant files had Utah *copyright* statements in them,
but without the expanded *license* statements, the licensing status of
the source files was unclear.

This commit is intended to clear up that confusion.

Most Utah-copyrighted files in the Emulab source tree are distributed
under the terms of the Affero GNU General Public License, version 3
(AGPLv3).

Most Utah-copyrighted files related to ProtoGENI are distributed under
the terms of the GENI Public License, which is a BSD-like open-source
license.

Some Utah-copyrighted files in the Emulab source tree are distributed
under the terms of the GNU Lesser General Public License, version 2.1
(LGPL).

make sure nothing has changed.

(and vnodes). Each time a node is allocated to an experiment it gets a
new root password (using the node_attributes table). The watchdog has
a new section that resets the root password (defaults to hourly).  We
still using a common password in the image to avoid totally bricking
ourselves, but once a node boots into an experiment it gets a new root
password.

This prevents hundreds of nodes with the same password, and all of the
problems associated with that.

indexed by exptidx. I also got the last of the pid and pid,gid tables.

Quick mod to stop an info leak.

bootlogs were persisting after experiment termination, only being replaced
when a particular node failed during TBSETUP (and hence sent back a boot
log to be stashed).  This was leaking info such as project and experiment
names, uids, groups.

For now the bootlog is being cleared as nodes come in to an experiment
via experiment swapin (inside nalloc).  running sched_reload or
sched_reserve will also call nalloc if the node is free, hence clearing
the bootlog as well.

When doing a swapmod, nodes already reserved to the experiment are "moved"
(via update) to a holding reservation. Fine.  After assign runs, the old
nodes are moved back, but this time by an insert into the table, which
causes them to lose some fields that I do not want them to lose! This might
really mean that these fields do not belong in the reserved table, but I do
not want to ponder this right now. Instead I do another update bringing
them back into the original experiment.

I left comment indicating that this is under review (and why this commit is
seperate from the rest of the swapmod changes).

return value, and the return value for die() is not as predictable as
you might think!

out of the reserved table. Mostly this happens in nfree and nalloc,
but there a couple of other moves, in libdb and in the reload daemon.
The uid and experiment are stored, long with a timestamp.

* Put in debugging timestamps to figure out why so slow.

* Cleanup, fixing bad identing, bad style, bad comments, etc.

* Restructure so that instead of locking three tables (nodes,
  reserved, and node_activity), we just need to lock the reserved
  table, by moving the node existence check to before the main loop,
  and the node_activity update to after we unlock tables. Not locking
  the nodes table results in less serialization and less waiting for
  the lock to be granted.

* XXX: Check each node for an entry in tiplines table; Do not call
  into console_setup.in if none of the nodes have tiplines. This is
  not a very clean thing to do of course, but it saves up to half the
  time it takes to do an nalloc cause of the min startup time for our
  perl scripts (almost 1/2 a second). When allocating virtnodes on a
  physnode, no reason to even call console_setup.in.

Bottom line? Well, not as good as one would like ... time to allocate
10 vnodes on a pnode goes from .8 seconds to .4 seconds ...

from commvirtsig-branch into main trunk

Two testsuite tests fail: tbcmd and vtypes . Need to fix.

nodes it can).

Change exit value; return -1 on fatal error, otherwise return the
number of nodes that could not be allocated. Combined with the -p
switch, assign_wrapper can easily determine that nalloc was able to
reserve a subset of the nodes.

Also fix up getopts() call, which had its arguments backwards! Good
thing we hardly pass switches to nalloc.

real life.

There's a small window in time (typically a few seconds, but maybe 10's of
seconds on occasion) between the time nodes get nalloc'd and the time they
get rebooted during a swapin. If idlemail runs (or someone views a web
page) during this time, the expt can appear idle because the nodes haven't
had any activity. It can cause an incorrect email or a false positive on
the web page.

The fix is to have nalloc register some external node activity, like
node_reboot does, when it allocates a node.

reserved table. Whenever you insert a reservation or change a reservation
to a new expt, set vname=node_id.

what it does.

Capserver and capture now handshake the owner/group of the tipline.
Owner is defaults to root, and the group defaults to root when the
node is not allocated. Capture will do the chmod after the handshake,
so if boss is down when capture starts, the acl/run file will get 0,0,
but will get the proper owner/group later after its able to handshake.
As a result, console_setup.proxy was trimmed down and cleaned up a
bit, since it no longer has to muck with some of this stuff.

A second change was to support multiple tiplines per node. I have
modified the tiplines table as such:

	| Field   | Type        | Null | Key | Default | Extra |
	+---------+-------------+------+-----+---------+-------+
	| tipname | varchar(32) |      | PRI |         |       |
	| node_id | varchar(10) |      |     |         |       |
	| server  | varchar(64) |      |     |         |       |

That is, the name of the tip device (given to capture) is the unique
key, and there can be multiple tiplines associated with each node.
console_setup now uses the tiplines table to determine what tiplines
need to be reset; used to be just the name of the node_id passed into
console_setup. Conversely, capserver uses the tipname to map back to
the node_id, so that it can get the owner/group from the reserved
table.

I also removed the shark hack from nalloc, nfree, and console_reset,
since there is no longer any need for that; this can be described
completely now with tiplines table entries. If we ever bring the
sharks back, we will need to generate new entries. Hah!

standardize on using /usr/bin/perl instead of /usr/local/bin/perl (most already
did.)

we don't have a root user in the DB. Perhaps this is bad?

re-writing and formatting changes to make them more maintainable.  Also
re-worked command line handling to use Getopt::Std.

The problem is that "@TESTMODE@" is wrong! This becomes a string "0",
instead of the integer 0, and so "if ($TESTMODE)" breaks down.

Rob and I noticed that when we need to do maintainance to a node, its hard
to grab it when it gets freed before it gets taken again. So this will allow
admins to set up nodes so that when the current owner frees them, they
immediately are moved into another experiment. For instance, if I need to do
an upgrade on all the machines in the testbed, I can do:

sched_reserve testbed down pc1 pc2 pc3 [...] pc40

and it will immediately reserve any free nodes, skip any that are already
in testbed/down, and add an entry to the next_reserve table for any nodes
that are already reserved. Then nfree checks for entries in the next_reserve
table before it lets any nodes go (much like it does for the reloads table),
and moves them into the expt before it can get reserved again. So we can
guarantee that the next time a node is freed, it will be saved for us.

Note that this can also be useful to accumulate nodes for a large experiment,
since you can make sure you get them as soon as the current owner is done.

machines in the experiment. exports_setup runs on paper and queries
the DB, generating a set of export lines. That is sent over to
plastic, and piped into exports_setup.proxy, which generates a new
/etc/exports file and restarts mountd. All of the client setup scripts
have been changed to mount the project and user directories.

list of nodes for console_setup to avoid an ssh per node. Change the
sleep(1) in console_setup.proxy to a select(,,,0.25). Looks to be
about .25 seconds per node now, plus a constant time of 1.5 seconds.