GitLab

the extend modal, tell the user they cannot extend their experiment
until the cluster is back online, try again later. This is a stopgap, we
probably need a better way to handle transient failure in contacting
clusters when doing extensions.

an experiment from the cluster (say, cause its offline).

1. Reservation system now groks experiment lockdown and swappable. When
   swapping in, lockdown and swappable mean the expected end of the
   experiment is never.

2. Reservation library now handles changes to lockdowm, swappable, and
   autoswap (timeout). editexp now hands these changes off to a new
   script called manage_expsettings, which can be called by hand since
   we might need to force a change (I am not changing the classic UI, if
   a change is not allowed by the res system, we have to do it by hand).

3. Minor fixes to reservation library.

It was printing out the destination (aka, boss) rather than the IP
that was trying to access boss. Very confusing.

with sufficient permission to do anything useful (including geni users).
Head them off at the pass and explain in simple (short), direct (short)
clear (short) text why they can't do what they want to do.

Note that node_type_attributes.bios_waittime could be used to
dynamically adjust the stated timeout, but I don't want to embed
semantics of a particular state in stated, so we would have to
have some more general mechanism to tell stated to adjust the
timeout value based on a database field.

getticket before the underlying experiment is created. This is a temp
fix for Hussam, better fix is under development in my tree.

checker) when the number of nodes is greater then 100. There appears to
be a pathological problem in Jacks that causes it to consume 100% of
your CPU for 5 minutes or so before finishing. The was making it
impossible for David to do his 200+ node tests.

Addendum: while testing this just now, I noticed that the problem is
related to the cross product of nodes and links. 100 nodes and no links
is actually not that bad, but add a lan and watch out!

for a more complete picture of what the user was seeing.

Can't live with em, can't kill em dead... When writing my hack
routine to convert an image path into an imageid, I failed to
consider the .ddz (delta image) suffix.

Added a couple of more recent ones: autofs and ixl driver.

Turns out some combination of ebtables userspace and kernel doesn't
respect the --stp-type matcher.  So just drop all forwarded packets
destined to the bridge group address.

Anyway, this STP-less firewall bridge should be a better fit for most
switches.

Given that in an Emulab per-experiment firewall, there is only one
switch port in the experiment that is in the default control net vlan
(the firewalled nodes' ports are only in the per-experiment private
control net vlan), there is no risk of a control net loop, so it is safe
to turn off STP for the firewall's control net bridge.

However, when STP is off, Linux then seems to forward BPDUs across the
bridge (i.e. https://lists.linuxfoundation.org/pipermail/bridge/2007-April/005406.html),
which we don't want.  They intended it to support transparent bridges,
but this is not a transparent bridge, and there is no risk of it causing
a loop scenario.

we can infer if an extension request has been limited by the reservation
schedule.

we can infer if an extension request has been limited by the reservation
schedule.

I noticed that group_root could not delete users from projects. Seems
like we should allow that, but with the restriction that a group_root
cannot delete another group_root. Simple enough, right? Well thats not
how the permission system works; permission to do stuff to users is
based on who you are in the project, not who you are doing it to.

And then there are the subtle differences in permission handling between
the Classic interface and the Portal interface. And I am fully
unmotivated to fix anything in the Classic interface, hard to believe?

Anyway, most people are not going to notice anything since the bulk of
the changes affect sub groups. Sigh.

interfaces from the wires table. Helpful for debugging.