Re: [Condor-users] Suspended --> Retiring transition

[Dan Bradley <dan@xxxxxxxxxxxx>]

Jimmy Choi wrote:

> Thanks Dan. Does the state stays in Suspended because the previous
> Retiring state is triggered by PREEMPT? What if we have:
>
> Time 1: Busy->Retiring (due to machine rank)
> Time 2: Retirng->Suspended
> Time 3: PREEMPT evaluates to true
>
> At time 4, will there be a Suspend->Retiring transition?
>
> My experiment tells me that the answer is yes, but it'll be great to
> confirm this.

Due to a subtlety of the design, yes, there will be another 
Suspended->Retiring transition in this case.  Internally, there are two 
kinds of retirement: reversible and irreversible.  PREEMPT causes a 
transition to irreversible retirement, while a preempting claim causes a 
transition to reversible retirement, so if the preempting claim goes 
away before the retirement is completed, the original claim backs out of 
retirement.

--Dan