VDT 1.3.7 was released on Tuesday. Lots of things changed, perhaps most notably was the addition of GT4. Complete release notes for VDT 1.3.7
VDT 1.3.8 is in progress, and hopefully it will be a short development cycle so we can release it before OSG 0.4 is released. You can look at our internal to-do list and tell us if we're missing anything that must be in VDT 1.3.8. Be warned: this page is often not up to date, so don't consider it to be accurate on any given day.
There was discussion about PRIMA for GT4's web services GRAM. Right now, it doesn't exist, and that will cause problems for many VDT users that wish to move to GT4 GRAM. Vikram Andem from Fermi is investigating this. It looks hard, unless GUMS supports SAML2 and we can use a Globus interface to directly communicate with SMAL2. If not, it may be a lot of work. Globus has some documentation about authorization callouts for GT4 web services.
Leigh asked about sudo and GT4 GRAM. sudo is required, but it doesn't
require system adminstrators to list in
the accounts that might be used to run jobs. Instead, you can allow
ALL, !root to allow all users that are not
root. This may allow more users than a system adminstrator wishes, but
will not require constant updates to
We noted that when sudo it used, it normally invokes the globus-gridmap-and-execute program, which looks up a user in the gridmap file, and will only run if the user is found there. This program cannot make any authorization callouts (as best Alain can tell), but it can be removed from here if desired: it's merely a double-check because the authorization was also performed earlier.
Greg Cross asked about detecting VDT_LOCATION or the OSG installation location on a server that is running a Globus not from the VDT, like a TeraGrid site. There is no easy way to do this. The VDT (hopefully 1.3.8) will propagate some environment variables to jobs so that they can know things like this, but it won't help on a site that isn't using Globus from the VDT.
Shaowen said that an update to the GIP configuration will be released soon. It will add a missing attribute and allow SRM information to be published. He'll give it to the VDT and Leigh in the next couple of days. It will be in VDT 1.3.8.
There is a new VOMS in the VDT 1.3.7 and a new VOMS Admin. They will both be updated again in VDT 1.3.8. VOMS in 1.3.7 still has the KCA problem experienced by the Fermi folks. We don't know when it will be fixed, but if it is it will be updated in the VDT shortly thereafter.
Greg asked for SuSE 9 support. He had to do a gross hack (add an
/etc/redhat-release file) to get the install to work with
-pretend-platform. Alain thinks that he doesn't
need to do this anymore with VDT 1.3.7. Although SuSE 9 isn't
officially supported, it should work okay with