GitLab

in the pipe. When invoked from the web interface STDERR has
been redirected to the apache log file, and if we do not reset
it, error output land there instead of in the pipe we just opened.

And add DEBUG setting in rc.d startup scripts. This is something we had been
doing on subbosses with a hack, but I kept overwriting the hack!

when created via the ProtoGeni interface; we run everything as
GeniUser and so lose that info. We want to know this for history
information and so we can manage permissions on image deletion
effectively.

Also add updater, updater_idx, and updater_urn to track last change
to the image.

CAVEAT: For system images, we do not actually overwrite the image, but
store it out in /proj/$pid/images and then sometime later maybe we
copy it back to boss. This wrecks the tracking info quite effectively
since the image may never be copied back. Needs more thought.

message from scripts that want to have their own.

demonstrate usage.

so that it can be called from the PG code.

Previously, all certificates were created without URNs, and ProtoGENI
sites bolted them on when initsite was run.

script and call it from initsite.

update-install cause Node is used before code install.

warning.

ops apache certificate.

swap in yet:

set client [$ns node]
tb-set-node-os $client FEDORA15-STD

set d1 [$ns blockstore]
$d1 set-class "SAN"
$d1 set-protocol "iSCSI"
$d1 set-size 2GB

set san [$ns duplex-link $client $d1 10Kbps 0ms DropTail]

we can determine precisely what blockstore is reserved under the
name of the pcvm. For each blockstore, there is a pseudo pcvm node
for it. Mostly this is for nfree, so that we can safely release a
a node and have the accounting done properly.

to support swapmod, since otherwise we would not know what an
experiment currently has once the swapmod starts and the virt topo
cleared from the DB.

Will be used to explicitly mark which blockstores should be made available to
other nodes.

Apparently, on a swap modify, RemoteDeleteVlan gets called with an array
of vlan names rather than vlan objects. Needed to fix a couple of places
in the ELABINELAB path where this was not considered (the "regular" path
does handle this case).

While everything appears to work correctly now (i.e., inner DB state is
correct, outer vlans are set up correctly on switch), someone who knows
the code should double check this!

project/group membership. Confirm this with a confirm page. This
will solve the problem of adding a brand new user to an admins only
project, before they are an admin.