GitLab

* Cleanup! A lot of the structure derived from the early frame days,
  which had a noticable (and bad) effect on how I wrote the stuff.  I
  cleaned up most of that yuckyness.

* In process, optimize a little bit on the queries. The old code did
  about 9 queries just to write out the menu options, and then
  repeated most of those queries again in the page guts. I've
  consolidated the queries as much as possible (to 3) and cache all
  the results.

* Fix up problem with users who forget their passwords before
  verification. Basically, I fixed the more general problem of not
  being able to update your user info before verification/approval;
  users now get that menu option no matter their status.

* Fix up problem of users being able to access pages before
  verification (but after approval) by going around the menu options.
  The page level check (after the menu is drawn) now checks all
  conditions (password expired, unverified, unapproved, timedout, and
  also nologins()).

* Minor change in approveuser; do not show the new account to the
  project leader until the new user has verified his account.

* Change verification method, as reqwuested by Dave.  In addition to
  providing the key, also provide a web link to take the user straight
  to verification. I actually take them direct to the login page, and
  pass the key in as an argument. If the user is already logged in,
  bypass and go directly to the verify page (not the form page of
  course).  If the user is not logged in, let him log in, and then
  forward the key onward to the verify page. Basically, bypass the
  form all the time, and just do the verification.

* Minor change in showuser; Do not show pid/groups not approved in,
  and if the count is zero, do not draw the table headings.

database (iddata_clean), and a script to generate fake idle entries for
testing scripts on the minibed (idloadup.pl).

iddata_clean, by default, removes all idle entries older than 36 hours.
This can be changed by providing -m <maxage> (in seconds) to it.

idloadup.pl adds random idle data at time offsets from now, to 5 days
prior (1 sec granularity).  It adds anywhere from 1 to 10 entries per
PC to the node_idlestats page, and from 1 - 10 x # interfaces to
the iface_counters table.

of "fast insert" mode. That option plus a couple of others makes
duplicating the DB 3-4x faster.

Add a DB table of tables to prune out of the copy before creating the
clean DB. There is a lot of stuff in the DB that is not needed for the
testsuite, and the constant copies churn the DB for little reason.

various redundant spots though.

return value on VLAN creation. If it reports success, we look for the
VLAN with findVlan(), to make sure it's really there. We retry a
couple times if it didn't work. This will add a small (sub-second)
time penalty per VLAN created.

We ran into one instance, that we've noticed, where the switch
reported that VLAN creation succeeded, but the VLAN did not actually
get created on the switch.

Change default idledays from 3 to 2, and don't show the swap request button if the expt is marked unswappable. (I'll mark things unswappable as necessary for special expts, even if the original creator doesn't.

size. Works nicely!

again. No biggie.

(Added for my timing tests.)

limited to 31 characters, so we can't give commands with more than 8
outlets. So, we split up the outlets into groups of 8, and issue
multiple commands, if we are passed too many outlets.

Took some minor re-structuring, since we now may need to sync up with
the RPC prompt more than once.

a new hack for remote nodes! All remote node solving is done in the
new wanassign script. That script deals with both phys (ron) and
virtual (vrons), as well as fixed nodes of either type. The solution
is returned to assign_wrapper, which still does the tunnel stuff.
At some point it would be nice to get general purpose virtual nodes
(multiplex) working on either local or remote nodes, but thats a ways
off cause it means too many changes in assign_wrapper, which really
needs a total rewrite.

tb- command passed along in the DB). Add check to make sure no mixed
use of types and classes since. Change solver invocation. Was trying
to use IPC goo to pipe in input to solver and read back results,
abd tee off the input file for debugging. That does not work cause the
tee never exits, and so the child never exits. Reorg slightly; write
the input file to a temp file and pipe in later. I need to kill off
the IPC goo though, since its no longer needed, but its harmless as it
is.

used when the virtual topology does not include any link to the
testbed, but consists only of remote links.

running in "impotent" mode; just make them up.

compatibility.

longer used.

Changes for Leigh: Force active view to show all expts that are active or activating. Also, count ron nodes as PCs.

Also, last revision included Mike's gated change in the linux version.

"prepared", so it was trying to delete the testbed groups/accounts.

from Dave.

that subnets are actually unique!

Also add required changes to wanlinkinfo and wanassign.

with cvsup.

Controls whether we user recent or hysterical widearea_delays data.
Defaults to 0.

Tee off the wanlinkinfo output into a log file for inspection.

have a remote node as a member. Since we share remotes nodes and set
up tunnels, we cannot let two experiments have the same subnet
numbers. I use 10.0.0 nets and a DB table to allocate them with a
autoincrement field.

remote nodes so that we do not overlap between experiments.

This script grabs updated latency from Dave Andersen't database
of RON information, and sticks it into the widearea_recent table.

placing a "-1" anywhere in the latency or bandwidth matrices will result in
huge penalties when that connection is used in the mapping.

Also added neat (possibly novel?)
"mutation spike" which, when the standard deviation of the population gets
beneath a certain value, spikes the mutation rate from 1% to 50% for a generation,
thus keeping the deviation (and the population diversity) up.
Results from testing look promising.. in my "standard" 6-node mapping,
the solver finds the "best" solution in 95% of the runs.
(vs about 50% or less the old way.)