:) Speechless ... I will stress test it tomorrow!

Am 22.04.2014 21:24, schrieb John Chilton:
On Tue, Apr 22, 2014 at 1:59 PM, Björn Grüning
<bjoern.gruen...@gmail.com> wrote:
Hi John,

Am 22.04.2014 20:45, schrieb John Chilton:

You can create multiple variants of the 10G_memory tool with different
native specifications right? Does this work around result in too much
XML? Dynamic destinations could also be used - if you are interested I
could post a demo rule that you might be able to use.

You are right of course - very sorry. Dynamic destinations and macros
would make sense if there was some way to alter the environment based
on the destination - which I am not sure is possible with
nativeSpecification at least generally. (It sounds like you can with
PBS/Torque).

Here is patch that would add the ability to specify <env
id="NAME">VALUE</env> tags to job destinations in job_conf.xml and
have them auto-populated in jobs.

https://github.com/jmchilton/galaxy-central/compare/job_env?expand=1
https://github.com/jmchilton/galaxy-central/commit/d05e15d712c9a3175552e7c5f7041301507dae1f.patch

Any interest in testing it for me :)?

If not, I can open a PR for it as it is.



The problem with multiple variants (multiple handlers) is that I only can
specify one ENV variable that is known to the current shell. I don't think I
can set the variable directly inside the nativeSpecification tag, or?
Moreover, the variable need to have the name _JAVA_OPTIONS.
To solve that I need to set the variable somehow inside the
nativeSpecification tag.

Dynamic destinations may work, but what happens if I change a ENV inside of
python ... it is not known to the parent process or? Even if it is known, is
it only known to that specific tool?


The memory card referenced above might help, but a much easier cards
might help this would be...

https://trello.com/c/t1FH1Q5P



Yes, but that is really hackish :). If there is no other way, I will try
something like that ...


Also tool style macros for job_conf.xml might help right - keep size
of XML down?

https://trello.com/c/wt1LflNh


Unfortunately, macros do not work here. In GATK different tools do have
different requirements.
I can probationally work around that with HEAP_SIZE/2 or something like
that, but a job_conf.xml based solution would be better.


Feel free to create another card for this request (allowing
parameterized tools to influence destination parameters) - seems
reasonable.


Ok I will wait a few more hours and than create such a card.
Thanks for your ideas,
Bjoern





-John

On Tue, Apr 22, 2014 at 1:32 PM, Björn Grüning
<bjoern.gruen...@gmail.com> wrote:

Hi,

to specify -Xmx and -Xms for java based programs I set my preferred
values
to $_JAVA_OPTIONS and include _JAVA_OPTIONS as nativeSpecification with
-v
_JAVA_OPTIONS. That worked good so far, but the downside is that I can
not
have different options for different tools. Is there a trick I can
specify
arbitrary $ENVs to a specific tool? Any ideas how to solve that, without
hacking the -Xmx and -Xms parameters into the wrapper file?

It is related to https://trello.com/c/HGEYPQf6.
I would be interested in workarounds until that card is closed :)

Something like that would be great:
<tool id="toolshed.g2.bx.psu.edu/repos/iuc/gatk2/gatk2_indel_realigner"
destination="10G_memory">
<param id="nativeParam">export _JAVA_OPTIONS='-Xmx=6GB'</param>
</tool>

Thanks!
Bjoern



___________________________________________________________
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/
___________________________________________________________
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/

Reply via email to