[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Condor-users] ClassAds of hibernating machines

> Is a "sleeping" machine any machine with which an earlier
> established connection has been lost? Or are "sleeping"
> machines only those who have been hibernated by Condor?
> In other words:
> is, or is there not a difference between hibernated machines,
> and machines who got cut off otherwise (disfunctional internet
> connection; sudden power off; etc.).

Yes, there is a difference. A disconnected machine is simply just gone
once the ad expires on the Collector; hibernating machines, on the other
hand, are persisted by the Collector.  It is all stored in the original
machine class ad being held on the Collector (with some added attributes
to help revive it, like the hardware address, subnet mask, etc).  This ad
can sit on the Collector indefinitely, or it can be made to expire (as in
the case when a hibernating machine is physically removed from the pool,
and does not respond to wake-up calls).

So to answer your first question, "sleeping" machines are ones that have
entered hibernation because of Condor.  They do so via a user defined
policy: for instance, a basic policy might be to say that when all of the
machine's slots have been idle for 2 hours, the machine should change to a
different power state, like a Windows "sleep" or soft off, etc..