GitLab

PLABSUPPORT=0.

experiment "start" (initial creation with swapin). Been this way for a
few months, but the fix to the DB was easy:

	update experiment_resources as r
	left join testbed_stats as s on s.rsrcidx=r.idx
	 set swapin_time=UNIX_TIMESTAMP(tstamp)
        where action='start' and exitcode=0 and swapin_time=0;

dropped out of the tally. Also break out swapmods (active vs inactive)
and preload vs starts.

- latency measurements under 2 second intervals now operate in "continuous"
  mode: a single "ping" process runs per destination, versus the "normal"
  approach of forking a new fping process for each measurement instance.

- Note, that I found a bug which is not fixed in this checkin: if a latency
  test is running in continuous mode and another latency measurement request
  is received with a shorter interval, the "continuous" ping process is not
  re-initialized with the new, shorter interval.

changes are on the client side where I took mkjail and retargeted it
to vservers (called it mkvserver.pl, clever eh?) in the linux
directory. The real time sync was understanding how vservers work, how
they boot how they die, how they handle signals, etc. Very interesting
and very bizarre. Anyway, this first cut is done with the version 2.2
vserver code which does not virtualize the network stack or even the
loopback device, so I pretty much ignored the experimental network and
the host routine stuff. So, in your NS file you can now do this:

	set ns [new Simulator]
	set v0 [$ns node]
	set v1 [$ns node]

	tb-set-hardware $v0 pcvm
	tb-set-hardware $v1 pcvm
	tb-set-node-os $v0 FC-VSERVER
	tb-set-node-os $v1 FC-VSERVER

As you can see, I am using the osid to indicate jails vs
vservers. There are some small changes in assign_wrapper that use the
nextosid of the osid to map to the actual osid to install on the
hosting node. If you try to collocate a jail and a vserver assign will
refuse, cause we use features and desires for the osids. Sweet.

Oh, the ssh button in the web interface does not work yet cause the page
assumes that local virtnodes can bind to port 22 in each vserver, but
that will not work yet.

between jails and vservers (freebsd vs linux). Temporary.

in the nodes table if there is no match in the interfaces table. This
is probably temporary until we get the vserver networking straightened
out. Unlike jails, packets from vservers actually have the IP of the
vserver as the src, not the physical host, but that does not
correspond to anything in the interfaces table. This change lets me
test things as they are now.

ez path which uses the same new id for both.

(and thus were confusing the older me...)

a nodeid.  rmanage is much more convenient to use for debugging than the
ASF libreboot wrapper...

Also add LOOPBACKMOUNT definition.

since that code uses one of those vars

tunnels.

This includes extending a crude hack to ensure that vnodes are only
  booted on certain types of vnodes (local pcs and homenet)

reflect this.

should not count against the global capacity limits.

attribute in the types table to cap the number of allowed virtual
nodes on each physnode (by type). We have to look in the reserved
table to see how many already allocated on each physnode, and adjust
the counts accordingly. Doing it here (via assign_wrapper) is not
atomic and so it could fail to catch an over subscription, but on a
lightly used testbed this will not be a problem in practice.

that is not the project the experiment is being created in. This
avoids problems when creating the inner boss since those subgroups
will not be present on the inner boss.

the state was backed up but we do not know if assign_wrapper removed
the state before it failed (with recover on).

A couple of minor fixes to other scripts noticed while debugging the above.

and " since the value never goes near a query without being sanitized.

would consider private, in particular IP addresses/DNS names.  Could
use more work.

revision ...