LSC Data Analysis Software Working Groups

Navigation

DASWG
LSC
LIGO


DASWG LAL Doxygen

Docs

How-to
Minutes
Technical
Software Docs

Download

Browse CVS
Repositories

Participate

Change Control Board
Edit these pages
Sub-committees
Mailing List
Telecon

Projects

DMT
geopp
Glue
LAL Home Page
LALApps Home Page
LDAS
LDG Client/Server
LDM
LDR
LIGOtools
MatApps
Metaio
Onasys
Online
OSG-LIGO

Agenda and Minutes, 9 Nov 2005

Agenda: Wed 09 Nov 2005 10:00 PDT


Today's DASWG Telecon

  a. SegFind Database Performance

  b. Onasys Performance

  c. Ganglia at the Sites

  d. Plieides Cluster

  e. RDS Generation

  f. Data Publication

  g. Software Release Status

Minutes

Attendance: 
	Kipp Cannon
	Shourov Chatterji
	Ben Johnson
	Ed Maros
	John McNabb
	Brian Moe
	Peter Shawhan


a) SegFind Database Performance

   Duncan -
        suggests that one doesn't query for more than 6 hours
        of segments at once.

        Apparently this helps SegFind queries.

   Ben -
        o load on SegFind DB is now low.
        o created a persistent DB conection for publishing
          which has increased his throughput without apparently
          damaging queries.


b) Onasys Performance

   Users of onasays are:
        Kipp    : Excess power
			All current failures are the result of dagman processes
			being SIGKILLed.
	Duncan  : Inspiral
        Xavi    : Strain Frames
        Lindy   : kleineWelle
			Has had problems due to Condor which appear to be
			resolved.
        Mendell : Generating SFT's:
			Greg is stuck because the database thinks he has too
			many processes running.  This is probably a bug in
			onasys.

   Onasys jobs, when terminated abnormally, do not update
   properly in the database.  This must be corrected.


c) Ganglia is Down At the Two Sites

   An upgrade was not quite backward compatible with the previous
   config files.  This problem should be solved by the end of the
   week. (Ben)

d) Pleides Cluster's High Activity

   There seems to be a concern that the Pleides Cluster had
   a high usage yet a LIGO user could not get any cycles.

   John McNabb: there are difficulties identifying the users by
      their IDs.  There is no general way of identifying users
      IDs and publically showing who is using the cluster without
      a policy change.

   Shourov: would like to know why his jobs are not running

   JM: Part of the the problem is that there are no run-time
      estimates that come along with the job.  There is no
      standard Globus way to do this.

   S: Yet I can run my jobs on the PSU cluster.

e) RDS Generation

   RDS generation is going well.  There was a problem with some
   channel renaming, but that apparently has been resolved.

f) Data Publishing

   Data Publishing has been going well.

g) Status of Software Releases

   As Planned.  Brian requested GraphViz be installed on gateway.

Adjourned @ 13:00 central time.
$Id: 051109.html,v 1.2 2005/12/07 18:12:38 patrick Exp $