On Thu, 6 Mar 2003, Ian Stokes-Rees wrote:
> Out of curiosity, does this mean Andrew has some long running backgrounded
> job with a one year grid-proxy?
>
> If not, how does the map "automatically" and "continuously" authenticate
> itself at the various nodes it contacts?
There's a roughly week-long proxy on the UI machine that does the job
submission ( grid-proxy-init -hours 168 -out /tmp/longproxy ) and the job
script derives a shorter lived proxy from that when it runs
( grid-proxy-init -cert /tmp/longproxy -key /tmp/longproxy -hours 24 )
Incidently, the wget callback to the website that is the measure of
success for both direct Globus and indirect brokered jobs is done on port
2811 (even though the protocol is actually HTTP.) So unless people only
allow outgoing GridFTP from their worker nodes to certain sites, then
there shouldn't be a problem. (And if there is, then both the Globus and
RB jobs would fail.)
Andrew
|