> Date: Mon, 23 Jun 2014 22:19:27 -0500
> Subject: Re: [galaxy-dev] Run a tool's workflow one by one ?
> From: jmchilton@gmail.com
> To: leonardsquall15@hotmail.com
> CC: galaxy-dev@lists.bx.psu.edu
>
> Its looks like you were going to post an error message but didn't.
> That might help debug problem.
>
> There is no way currently at the workflow level to force one job to
> wait for another before completion (other than assigning an explicit
> input/output relationship between the steps). There is a Trello card
> for this here https://trello.com/c/h5qZlgU8.
>
> I am not sure that Trello card is really the best approach for this
> problem though. If it really is the case that these jobs can run
> simultaneously and they are not implicitly dependent on each other in
> some way not represented in the workflow - than it is likely they are
> running on a machine that just doesn't have enough resources (likely
> memory) to run these properly. The correct solution for this I think
> should be properly configuring a job_conf.xml file to not let Galaxy
> tools over consume memory.
>
> By default Galaxy will run 4 jobs simultaneously - any job of any time
> - regardless of memory consumption, threads used, etc.... This gist
> (https://gist.github.com/jmchilton/ff186b01d51d401623be) contains a
> job_conf.xml that you can stick in your Galaxy root directory to
> ensure a handful of tools (I used ids hilbert, fft, slm as example ids
> but you should replace these values with actual values of your tool)
> can only run one job at a time. All other jobs will continue t
> concurrently run two at a time beside these.
>
> If you are using a distributed resource manager (like sun grid engine,
> SLURM, Condor, etc...) then the solution is a little different. You
> should assign these tools to job destination that consume a whole node
> - you would to provide more information about the cluster hardware and
> software configuration for me to provide an example of this.
>
> Beside that the common advice about scaling up Galaxy holds - you
> should configure Postgres instead of sqlite, setup a proxy (nginx or
> Apache), disable debug in universe_wsgi.ini, etc.... See
> https://wiki.galaxyproject.org/Admin/Config/Performance/ProductionServer
> for details. All of these things can help in situations like this.
>
> -John
>
>
>
>
> On Mon, Jun 23, 2014 at 11:22 AM, Pat-74100 <leonardsquall15@hotmail.com> wrote:
> > Hi !
> >
> > I have a big workflow and sometime when I launched it, I got an error
> > message for some of my tools : unable to finish job.
> > I think it's maybe Galaxy ran multiple job so I get this error message.
> >
> > I'm looking to run my workflow "step by step".
> >
> > For example this workflow:
> >
> > http://hcsvlab.org.au/wp-content/uploads/2014/02/PsySoundTest1.png
> >
> > I'm looking to run Hilbert THEN FFT THEN SLM and no Hilbert, FFT and SLM at
> > the same time.
> >
> > Is it possible to make a workflow which wait to finish a job before run an
> > another job ?
> >
> > Thanks
> >
> > ___________________________________________________________
> > Please keep all replies on the list by using "reply all"
> > in your mail client. To manage your subscriptions to this
> > and other Galaxy lists, please use the interface at:
> > http://lists.bx.psu.edu/
> >
> > To search Galaxy mailing lists use the unified search at:
> > http://galaxyproject.org/search/mailinglists/