Software & computing meeting 20/07/2010
by
Marie Anne Bizouard
—
last modified
2010-07-21 16:41
********************************
Announcements:
********************************
- agreement to set up a new software&computing mailing list open to LSC colleagues
-->Action item for MAB
- rawdata rate:
Antonella: 8 MB/s --> circular buffer size: 212 days
10 MB/s --> 169 days
11.3 MB/s --> 150 days
--> people should be aware of this reduction. + extra cost in CC
Alain: Step 11 data rate seems to be around 11 MB/s. Need to wait for few more
days before deciding. Rate changes according to ITF step.
The new channels are important for commissioning. What is stored in CC could also be reduced.
Rate web page: http://wwwcascina.virgo.infn.it/MonitoringWeb/DAQ/index.html
Antonella: money issues.
Action items: need to check disk space in Bologna (35 TB only available now). What
has been asked for 2010 is not yet available.
--> Action item for Gianluca: will resend an email to Fulvio
- transfer:
No major concerns, but need to check Bologna disk space situation.
*******************************************************
1. Virgo/LSC software installation
*******************************************************
Franco: 2 basic problems:
a/ OS + compiler: Virgo SL 4 + 5 (32 & 64 bits)
LSC: CentOS 5 + Debian 5 + (Fedora + .... supported on best effort)
CentOS is the closest OS to SL.
LSC will soon form a committee to decide on next reference platform (possibly CentOS6 and Debian Squeeze)
Adam mentions that there is a working group thinking about OS support in the future.
Which platform for Adv Detector area? Virgo has to be involved in these discussions
Franco: to help installation one needs to check that everything goes right on each OS. Need a test machine
at cascina. Virtual machine with different OS.
b/ Software build:
Virgo uses CMT. LSC uses autoconf tools. Needs compromise to be able to install
easily both LSC and Virgo software on both side.
Adam and Franco started discussions. Should be able to use CMT to install native Virgo package in LIGO
clusters (at least initially, would finally like to get RPMs/Debs packaged from CMT build - LIGO issue). Waht is then needed by autoconf is a file (pkg_config) specifing where are the libraries and
the include files
ACTION item for Franco and Adam: see how this file can be automatically generated at building time.
c/ exchange: cvs and git.
rpms: would require pre compiled lib. + pb of root privileges in CC. --> not recommended.
LSC package used in CMT: needs a CMT wrapper package. Already exist for lal/lalapps. see mbta f.i.
d/ Specific problems encountered by users:
- Nicolas mentions DMT installation: require lots of ext package. Nightmare
- Gianluca had pb with lal/lalapps at Bologna. Mainly because of ext. packages dependency.
- Require to install lsc-soft. + pb with condor 32 bist vs 64 bits
Action item for Adam: provide the DASWG page where all list of ext packages are listed:
--> https://www.lsc-group.phys.uwm.edu/daswg/download/repositories.html
Adam mentions that condor team is thinking to make compilation free of condor libs.
**********************************************************
2. Condor job flocking
**********************************************************
Project mean to test the facility to run cbc pipeline in Virgo CCs.
Important test for the future: is that the right solution for Adv detector area pipelines?
On-going tests/projects:
- cbc follow-up jobs (Romain): require to access Virgo raw data. Cm based solution failed.
Romain reported (VIR-0457A-10) that, jobs have been submitted from UWM. 10-20 jobs per dag
(not too stressful). Found pbs. Condor job flocking generated pbs for the other jobs as well.
Still under investigation. Romain is asking for help from "experts". Now something like 1% of the jobs
fail. Could be acceptable but this is not obvious how the failing rate will scale with the number
of submitted jobs?
Increase of load foreseen? A priori yes from burst online searches requesting omega scans.
ACTION item for MAB: contact Igor to know if he would be interested to submit condor job on Bologna?
Use of Lyon? a priori yes, as soon as Bologna is stabilized.
But request from expert help:
Who can help?
Gergely, scott, Ross. Who else?
Who is involved in the project:
Romain, Gergely, Gianluca, Ross, Scott, xxx?
- cbc low mass ihope pipeline (Gianluca): the idea is to be able to run ihope in Virgo CC. S6b low mass
and then make a comparison with the rerun done in LIGO.
Solution : local condor library installed in Bologna + condor scheduler installed in Budapest.
Gergely have been able to submit jobs.
Gianluca is now facing software installtion pb + OS upgrade at Bologna (64 bit condor lib version is now
required).
--> needs help from Gergely
- Other project: Alex Dietz expressed an interest in the past to be able to run GRB pipeline in Virgo CC.
Conclusion: good progresses but need more help. expecially from condor + condor flocking experts.
Follow up at the next meeting.
Announcements:
********************************
- agreement to set up a new software&computing mailing list open to LSC colleagues
-->Action item for MAB
- rawdata rate:
Antonella: 8 MB/s --> circular buffer size: 212 days
10 MB/s --> 169 days
11.3 MB/s --> 150 days
--> people should be aware of this reduction. + extra cost in CC
Alain: Step 11 data rate seems to be around 11 MB/s. Need to wait for few more
days before deciding. Rate changes according to ITF step.
The new channels are important for commissioning. What is stored in CC could also be reduced.
Rate web page: http://wwwcascina.virgo.infn.it/MonitoringWeb/DAQ/index.html
Antonella: money issues.
Action items: need to check disk space in Bologna (35 TB only available now). What
has been asked for 2010 is not yet available.
--> Action item for Gianluca: will resend an email to Fulvio
- transfer:
No major concerns, but need to check Bologna disk space situation.
*******************************************************
1. Virgo/LSC software installation
*******************************************************
Franco: 2 basic problems:
a/ OS + compiler: Virgo SL 4 + 5 (32 & 64 bits)
LSC: CentOS 5 + Debian 5 + (Fedora + .... supported on best effort)
CentOS is the closest OS to SL.
LSC will soon form a committee to decide on next reference platform (possibly CentOS6 and Debian Squeeze)
Adam mentions that there is a working group thinking about OS support in the future.
Which platform for Adv Detector area? Virgo has to be involved in these discussions
Franco: to help installation one needs to check that everything goes right on each OS. Need a test machine
at cascina. Virtual machine with different OS.
b/ Software build:
Virgo uses CMT. LSC uses autoconf tools. Needs compromise to be able to install
easily both LSC and Virgo software on both side.
Adam and Franco started discussions. Should be able to use CMT to install native Virgo package in LIGO
clusters (at least initially, would finally like to get RPMs/Debs packaged from CMT build - LIGO issue). Waht is then needed by autoconf is a file (pkg_config) specifing where are the libraries and
the include files
ACTION item for Franco and Adam: see how this file can be automatically generated at building time.
c/ exchange: cvs and git.
rpms: would require pre compiled lib. + pb of root privileges in CC. --> not recommended.
LSC package used in CMT: needs a CMT wrapper package. Already exist for lal/lalapps. see mbta f.i.
d/ Specific problems encountered by users:
- Nicolas mentions DMT installation: require lots of ext package. Nightmare
- Gianluca had pb with lal/lalapps at Bologna. Mainly because of ext. packages dependency.
- Require to install lsc-soft. + pb with condor 32 bist vs 64 bits
Action item for Adam: provide the DASWG page where all list of ext packages are listed:
--> https://www.lsc-group.phys.uwm.edu/daswg/download/repositories.html
Adam mentions that condor team is thinking to make compilation free of condor libs.
**********************************************************
2. Condor job flocking
**********************************************************
Project mean to test the facility to run cbc pipeline in Virgo CCs.
Important test for the future: is that the right solution for Adv detector area pipelines?
On-going tests/projects:
- cbc follow-up jobs (Romain): require to access Virgo raw data. Cm based solution failed.
Romain reported (VIR-0457A-10) that, jobs have been submitted from UWM. 10-20 jobs per dag
(not too stressful). Found pbs. Condor job flocking generated pbs for the other jobs as well.
Still under investigation. Romain is asking for help from "experts". Now something like 1% of the jobs
fail. Could be acceptable but this is not obvious how the failing rate will scale with the number
of submitted jobs?
Increase of load foreseen? A priori yes from burst online searches requesting omega scans.
ACTION item for MAB: contact Igor to know if he would be interested to submit condor job on Bologna?
Use of Lyon? a priori yes, as soon as Bologna is stabilized.
But request from expert help:
Who can help?
Gergely, scott, Ross. Who else?
Who is involved in the project:
Romain, Gergely, Gianluca, Ross, Scott, xxx?
- cbc low mass ihope pipeline (Gianluca): the idea is to be able to run ihope in Virgo CC. S6b low mass
and then make a comparison with the rerun done in LIGO.
Solution : local condor library installed in Bologna + condor scheduler installed in Budapest.
Gergely have been able to submit jobs.
Gianluca is now facing software installtion pb + OS upgrade at Bologna (64 bit condor lib version is now
required).
--> needs help from Gergely
- Other project: Alex Dietz expressed an interest in the past to be able to run GRB pipeline in Virgo CC.
Conclusion: good progresses but need more help. expecially from condor + condor flocking experts.
Follow up at the next meeting.