Hi Alessandra,
Are you running only recon, or you have DQMs as well.
In June, this summer, before starting the pass1 of 2016, we noticed that
dropping DQMs, will allow to have
Memory request to go down to 2.4Gb with a heap size = 1.4G (-Xmx1400m).
If you have DQMs as well, could you try tr without DQM, otherwise if it
is only recon, them yes it seems to require too much of memory.
Rafo
On 11/06/2017 07:10 AM, Alessandra Filippi wrote:
> Hi all,
> I noticed that with the latest hps-java versions many of my jobs
> requiring reconstruction + info extraction for alignment of 2015 data
> fail due to LSF memory usage limit (on SLAC batch system).
>
> Since the alignment part did not change, I suppose some more memory
> usage is required for the reconstruction, when compared to previous
> versions (this is strange, though, since these failures only happen
> for 2015 data, not for 2016 runs as far as I can see).
> Did something change in the latest versions (4.0 of hps-java) which
> could be the source of this memory increase?
> Which memory limit would you suggest to modify the 4 GB default to get
> job come to completion? (I could try random values but if you have a
> wiser estimation on the momery increase it could be more useful).
> Thanks, cheers
> Alessandra
>
> ########################################################################
> Use REPLY-ALL to reply to list
>
> To unsubscribe from the HPS-SOFTWARE list, click the following link:
> https://urldefense.proofpoint.com/v2/url?u=https-3A__listserv.slac.stanford.edu_cgi-2Dbin_wa-3FSUBED1-3DHPS-2DSOFTWARE-26A-3D1&d=DwIBAg&c=lz9TcOasaINaaC3U7FbMev2lsutwpI4--09aP8Lu18s&r=0HDJrGO9TZQTE97J9Abt2A&m=Qlq5ypjjXnoktaAFX469rRPmo3_ewGl89N0SMdUp9Eo&s=HQ9HxOf7Kjty6e9n0_CzVtpYLuCMqw6zz5L_8zL3cOs&e=
########################################################################
Use REPLY-ALL to reply to list
To unsubscribe from the HPS-SOFTWARE list, click the following link:
https://listserv.slac.stanford.edu/cgi-bin/wa?SUBED1=HPS-SOFTWARE&A=1
|