GitLab

into per-experiment databases on ops. Additional support for reconsituting
those databases back into temporary databases on ops, for post processing.

* This revision relies on the "snort" port (/usr/ports/security/snort)
  to read the pcap files and load them into a database. The schema is
  probably not ideal, but its better then nothing. See the file
  ops:/usr/local/share/examples/snort/create_mysql for the schema.

* For simplicity, I have hooked into loghole, which already had all
  the code for downloading the trace data. I added some new methods to
  the XMLRPC server for loghole to use, to get the users DB password
  and the name of the per-experiment database. There is a new slot in
  the traces table that indicates that the trace should be snorted to
  its DB. In case you forgot, at the end of a run or when the instance
  is swapped out, loghole is run to download the trace data.

* For reconsituting, there are lots of additions to opsdb_control and
  opsdb_control.proxy to create "temporary" databases and load them
  from a dump file that is stored in the archive. I've added a button
  to the Template Record page, inappropriately called "Analyze" since
  right now all it does is reconsitute the trace data into a DB on
  ops.

  Currently, the only indication of what has been done (the name of
  the DBs created on ops) is the log email that the user gets. A
  future project is tell the user this info in the web interface.

* To turn on database capturing of trace data, do this in your NS
  file:

	set link0 ...
	$link0 trace
	$link0 trace_snaplen 128
	$link0 trace_db 1

   the increase in snaplen is optional, but a good idea if you want
   snort to undertand more then just ip headers.

* Also some changes to the parser to allow plain experiments to take
  advantage of all this stuff. To simple get yourself a per-experiment
  DB, put this in your NS file:

	tb-set-dpdb 1

  however, anytime you turn trace_db on for a link or lan, you
  automatically get a per-experiment DB.

* To capture the trace data to the DB, you can run loghole by hand:

	loghole sync -s

  the -s option turns on the "post-process" phase of loghole.

DB was created, and a flag to indicate if the DB is transient (and
can be garbage collected).

then just a dump file.

Not yet tested.

O(n) reduction in tournament time.

Currently leaks memory, but that can be fixed easily.

create a new template (well, really a modify) from the current
swapped in experiment. This allows you to create a template, swap in
an instance, modify the datastore in the instance (which is a copy
of the datastore in the template), and then create a new template
using the datastore and nsfile from the instance. This is a new menu
item on the showexp page for the instance.

Also in this commit are fixes and improvements to the new navagation
bar that I recently added.

of the most recent child (modification). Better then passing this info
back in a print statement to be parsed.

Small bug fixes in cleanup in os_setup summary code.

Fix syntax error.

swapexp: The previous commit, witch added a message about the recovery
action when a swap-modify failed to the top of the email, did not
catch all of the possible cases.  Added the case when the experiment is
not swapped in.

os_setup: Refactored/rewrote os_setup error summary code.  Distinguish
the case when nodes fail to properly load the os and when the don't
boot after loading the os.

Fix bug in libosload when there is an image problem.  The code that checks
if there is an image problem is part of the code that looks up information
on the image.  The problem is that this info is catched so that the next
time this info is looked up the checks are not performed.  Thus, libosload
will attempt to load the image even though there is a problem with the
image.  This will often lead to obscure error messages.  Fixed by using a
special value for when there is a problem with the image.

Fix bug when an experiment has more than 1 image type to load.  Only the
results from one of the image loads will be properly recognized. The
others will be ignored.  This is because the the same file handle is being
used for all the children due to the use of a global file handle glob.
Fixed by changing the glob to a variable and using "new IO::Handle" to
create a new handle for each child.

- Fixed; other logic defects.

- if a node cannot receive ACK's due to firewalled UDP, oprecv will kill
  off bgmon on that node. To discover this, opsrecv detects when a node
  sends many duplicates of the same data to which opsrecv has sent ACKs
  about in the past.

- added parameterized send rate of buffered entries.
- ACK's are checked for a timestamp to prevent old ACKs of same index
  numbers to be interpreted as current.
- common routines have been moved to a .pm file.
- added 'die' API function to force bgmon to exit.

install on an existing image.  Implemented by doing a client-install into
an empty directory via DESTDIR.

pursuit of standards compliance.  Nowhere near complete,
but you've got to start somewhere.
(all tags should be lowercase, all values should be quoted,
including simple things like width="100%").

Don't use "no warnings 'uninitialized'" since that is a perl 5.6+ feature
and some are still using an ancient version of perl.

link.

Also add an Update method.

Change user that receives slice email from the head of the experiment to
the experiment swapper.

Users that create Planetlab experiments will now get email coming back via
the emulabman@emulab.net alias.  The technique is simple:  stash away the
experiment head uid in a file corresponding to the slice's name over in
ops:/etc/mail/plab-slice-addrs.  The silly procmail config file I've committed
here reads this file (if it exists) and forwards mail to the user's email addr
contained therein if the incoming message contains the name of their slice.
There is no install target for this procmail filter;
I hand installed it and hooked it in.