On 01/15/2013 10:38 AM, Stephen Burke wrote:
> But since the content of the site BDII is completely under the site's
> control that shouldn't be a problem. Stephen
Reading back through this thread, I think these would be the
requirements for changes to the current system. We want to minimize the
number of places where downtimes are declared, in order to avoid
conflict and reduce the risk or misinformation. Also, the status of a CE
must persist whether the CE is on or off. And the status must be fully
controllable by admin without flip/flopping while the system is
reconfigured. Ideally, we want the monitoring agents and the WMS or
submission agents to consult the same information, to avoid running jobs
on downed systems. And the system must support states to finely control
the use of the CE (i.e. in or out of production, operating or not etc.)
so that testing through the WMS is supported even while the CE is in
downtime. At present, the status quo (or workaround) seems to be this:
Declare the CE in downtime in GOCDB.
When downtime comes, remove the CE from the site bdii by hand.
Run glite-ce-disable-submission to drain the CE.
Do maintenance.
Bring up system.
Local tests.
Global tests (using -r option job so the WMS sends the job directly to
the CE).
Put the CE back into the site BDII.
Stop downtime.
Run glite-ce-enable-submission.
Without a "one stop shop" for downtimes (aka GOCDB), I guess I'll have
to do it that way from now on. Anybody see any other problems?
Steve
--
Steve Jones [log in to unmask]
System Administrator office: 220
High Energy Physics Division tel (int): 42334
Oliver Lodge Laboratory tel (ext): +44 (0)151 794 2334
University of Liverpool http://www.liv.ac.uk/physics/hep/
|