[Swift-devel] block allocation

[Mihael Hategan]

You are correct in your assumptions.

There is one coaster service instance per machine per swift JVM.

On Thu, 2009-05-28 at 17:11 -0500, Zhao Zhang wrote:
> Hi, Mihael
> 
> Here comes another question about the definition of "block 
> allocation".From my understanding, "block allocation" is that we grab 
> bunch of
> resources, then run workflowS on it without request the resource again.
> Say, we request 1K compute nodes from a cluster, then we run a dock6 
> workflow on it, after that we run a blast workflow on the same allocation.
> Is my understanding correct?
> 
> In the current coaster design, coaster seems deallocate the resources 
> after A workflow is done. So, next time when I tried to run another 
> workflow,
> I have to run it on a new allocation instead of the one we have already 
> had in hand. Or is there any options we could set in coaster and could have
> both the approaches? Thanks.
> 
> best
> zhao
> 
> Mihael Hategan wrote:
> > You should check whether blocks are properly de-allocated when all the
> > work is done.
> >
> > On Wed, 2009-05-27 at 18:10 -0500, Zhao Zhang wrote:
> >   
> >> Sorry, Mihael, I didn't get your last question.
> >>
> >> zhao
> >>
> >> Mihael Hategan wrote:
> >>     
> >>> On Wed, 2009-05-27 at 17:33 -0500, Zhao Zhang wrote:
> >>>   
> >>>       
> >>>> Mihael Hategan wrote:
> >>>>     
> >>>>         
> >>>>> On Wed, 2009-05-27 at 16:49 -0500, Zhao Zhang wrote:
> >>>>>   
> >>>>>       
> >>>>>           
> >>>>>> Mihael Hategan wrote:
> >>>>>>     
> >>>>>>         
> >>>>>>             
> >>>>>>> On Wed, 2009-05-27 at 16:31 -0500, Zhao Zhang wrote:
> >>>>>>>   
> >>>>>>>       
> >>>>>>>           
> >>>>>>>               
> >>>>>>>> Hi, Mihael
> >>>>>>>>
> >>>>>>>> I did a clean run of 100 scip jobs on ranger. (scip is a new application 
> >>>>>>>> from MCS).
> >>>>>>>>     
> >>>>>>>>         
> >>>>>>>>             
> >>>>>>>>                 
> >>>>>>> And?
> >>>>>>>   
> >>>>>>>       
> >>>>>>>           
> >>>>>>>               
> >>>>>> I need your help to make sure coaster is working as expected.
> >>>>>>     
> >>>>>>         
> >>>>>>             
> >>>>> Did the workflow finish successfully? I'll assume "yes", since you
> >>>>> didn't mention any errors, but it would be useful to state such things
> >>>>> from the start.
> >>>>>
> >>>>>   
> >>>>>       
> >>>>>           
> >>>> Yes, it was successful.
> >>>>     
> >>>>         
> >>> That's not a bad sign.
> >>>
> >>>   
> >>>       
> >>>>> As far as it working as expected, I have more confidence in the workings
> >>>>> of the algorithm itself than in the ancillary stuff. In other words, if
> >>>>> you don't see errors/restarted jobs or stack traces in the logs, it's
> >>>>> likely that things are fine.
> >>>>>
> >>>>>   
> >>>>>       
> >>>>>           
> >>>>>>>   
> >>>>>>>       
> >>>>>>>           
> >>>>>>>               
> >>>>>>>> The log is at /home/zzhang/scip/ranger-logs/coasters.log
> >>>>>>>>     
> >>>>>>>>         
> >>>>>>>>             
> >>>>>>>>                 
> >>>>>>> [hategan at communicado ~]$ less /home/zzhang/scip/ranger-logs/coasters.log
> >>>>>>> /home/zzhang/scip/ranger-logs/coasters.log: Permission denied
> >>>>>>>   
> >>>>>>>       
> >>>>>>>           
> >>>>>>>               
> >>>>>> try now.
> >>>>>>     
> >>>>>>         
> >>>>>>             
> >>>>> [hategan at communicado ~]$ date
> >>>>> Wed May 27 17:26:51 CDT 2009
> >>>>> [hategan at communicado ~]$ less /home/zzhang/scip/ranger-logs/coasters.log
> >>>>> /home/zzhang/scip/ranger-logs/coasters.log: Permission denied
> >>>>>
> >>>>>   
> >>>>>       
> >>>>>           
> >>>> sorry, try again.
> >>>>     
> >>>>         
> >>> Looks ok. There are some things that should probably be adjusted there,
> >>> but it looks reasonable. Where the queued/running jobs removed from the
> >>> queue when the workflow finished?
> >>>
> >>>
> >>>   
> >>>       
> >
> >
> >