GitLab

Change the email going to testbed-errors from:

  From: user
  To: user
  Cc: testbed-errors

to

  From: testbed-ops
  To: user
  Bcc: testbed-errors
  X-NetBed-Cc: testbed-errors

This should cause all replies to these message to go to testbed-ops
instead of testbed-errors.  The only thing is that you need to be
careful when replying since if you only reply to the sender than it
will go to testbed-ops only and not to the user.  I was also thinking
about changing the other swap* failures messages still going to
testbed-ops in a similar way but due to the reply issue for those on
testbed-ops I will hold off on that for now unless someone else thinks
it's a good idea.

The addition of the X-NetBed-Cc header is to make filtering the email
easier since the fact that it is going to testbed-errors will no
longer be in the header.  This header is also present in the swap*
failures messages still going to testbed-ops.

doesn't have to redo the layout after they load.

installing a second one.

Fixed a time-calculation bug in the monitor. Added bandwidth and base-rtt calculations based on TCP-vegas. Miscellaneous cleanups.

TODOs

Don't start the plab renew daemon anymore.

Fix up plab renewal for current scheme.  We must now go out to each sliver
individually and ask for it to be renewed.  I've added a new script to be
run out of cron that will run through and attempt to renew all active slices.
If a node cannot be renewed with its slice and comes sufficiently close to
it's recorded expiration (currently two weeks out), mail will be sent to
testbed ops warning about this situation.  Note that there will only be
one email message per slice, containing a list of all nodes at risk for
expiration.  The plab renewal daemon will no longer run as a result of
this change.  Note that this is sort of a hack.  The better way would be
to have the daemon perisistenly try to renew nodes that have failed until
success, but that will take more work, and I might farm it off to the
plab monitoring daemon anyway.

Hhave swapexp/batchexp dump the error when the -w" option is
specified.  The error will look something like:

  ERROR:: <cause desc>

  <text of the error>

  Cause: <cause>
  Confidence: <confidence>

This will be the last thing printed.  The "::" is there to make
recognizing the error easy to scripts since they can just look for the
"ERROR::".

Add "plabeventkeys" tmcc command to libtmcc.pm

* Add Princeton IP subnet to the allowed list for viewing plabstats.php3
* Mention PlanetLab AUP on EZ creation page.
* Change language a bit RE PlanetLab AUP on the Portal doc page.

Change @TBBASE@ to @TBDOCBASE@ when refering to KB entry.

"firewallable" node types.

This way we can specify node_type_features for types that can be firewalled.

Right now that means that they can be powered off, which pcwifi's cannot be.

Fix up botched configure[.in].

command-line flag.

numbers off.  This should realign things...

Added libtblog.pm to configure

Fix bug causing strange errors from snipit due to an invalid assumtion
about __DIE__ handler in libtblog.pm.in.

Changed format of email sent to user on errors.  The error will now
appear instead of the generic message when I am confident it is
accurate.  The subject line will also change to reflect the cause of
an error.

Avoid sending mail to testbed-ops during failed swap related evenets
in some cases.  It will instead be sent to a new mailing list
testbed-errors.

Added a new row in the experiment info table "Last Error:" which
states the cause of the error, and links to a new page displaying the
error.

Made some assign/assign_wrapper errors more informative.

The error (as determined by tblog) is now stored in the database in a
more structured fashion.  This inlcudes adding a column for the session
(in the log table) to testbed_stats to link eash swap event with the
logs and possible the error.

Other changes to the database, see sql/database-migrate.txt

up a tcpdump on both the inside and outside interfaces on the firewall
at boot time.  Another useful form of debugging.

Note: as with the "accept" and "deny" styles of logging, "tcpdump" cannot
be set through any interface other than setting it in the virt_firewalls
DB table directly.

failed because of an NFS hiccup.  Also added a little activity
display.

side of iperf. The main goal is to use much less memory than
'iperf -s'.

Supports many of the same same command line args, so that it can be
used as a dropin replacement for iperf.

Currently only uses about 40k of memory, regardless of how many
clients are connected.

Seems to behave exactly the same as an iperf server, but more testing
is needed.

type.

Fixed a transient error discovered by Junxing. Rather, I've fixed a bug that could account for it and the error hasn't resurfaced after many runs. The transient symptom was when pcap only captured the first three packets and failed from then on.

All measurement times are now based on packet timestamps rather than gettimeofday(), throughput measurement is much more consistent now, though a more thorough evaluation is still required.

Delay is averaged over a quantum. This is probably not the right thing to do, but it is better than the alternative (latest measurement). I am still discussion with Sneha about what to do here.

layout correctly.

As it currently stands, they cannot be automatically swapped because
a paniced experiment will have a firewall node with its control interface
disabled.