[saga-rg] Clean Up after jobs - issue 61

[Christopher Smith]

Ultimately the job of cleaning up is left to the implementation, but it does
needs hints through the job description. I believe that JSDL has a richer
set of semantics for describing this kind of stuff than we currently support
through SAGA. Should we just map the JSDL descriptions of file operations
through to SAGA, or consider a minimal subset.

This kind of cleanup is most often (practically) done using post execution
scripts and the like within the queuing system, if not done explicitly by
the job script itself, rather than supported as an operation of a resource
manager. I don't know about brokering middleware and the like.

-- Chris


On 21/2/06 09:41, "Andre Merzky" <andre at merzky.net> wrote:

> Outside the SAGA use cases, we got feedback from some
> medical people about the job API.  Open questions in the
> discussion have been:
> 
>   - When are files created by a job cleaned up?
>   - Can we call clean_up?
>   - Can specify files to clean up, if default is to leave
>     them be?
> 
> Their concern is mostly privacy: they want to ensure that no
> data a job worked on are left behind.
> 
> Does anybody have insight into the semantics of the BES
> clean-up stage, or of other job managers?  Do other people
> also see the need for explicit clean-up?
> 
> If not (e.g. no answer in one week), I would leave that
> issue for later consideration, as we get more specific use
> cases in that respect.
> 
> Cheers, Andre.