Last Updated on Monday, January 14, 2002 at 12:06:07. by Author: Burt Holzman

Data Mover, General Information/Help

(Expert)

Contents

• Data Mover Overview
• The daq_daemon
• The ped_daemon
• The deletor_daemon
• The pedestal selector
• Why isn't the datamover sinking my sequence / running peds?
• How to contact author

Data Mover Overview

The data mover consists of three components: daq_daemon, which runs on phobosx; deletor_daemon, which runs on vmesparc; and ped_daemon, which runs on phast.  A fourth component, the pedestal selector, allows the shift crew to decide how often to run pedestals.  All four are visible on phucsia.

The daq_daemon

daq_daemon runs in a window on phucsia (which has been ssh'ed to phobosx), and the output looks like the following:

source: daq_daemon       , state: SEND_TO_HPSS,     filename: /foo/bar/PhoRaw000021s003.root
source: ped_daemon       , state: RUNNING_PEDS ,    filename: /foo/bar/PhoRaw000021s000.root
source: deletor_daemon  , state: UNKNOWN,              filename: 

If it crashes, CALL BURT.  

The ped_daemon

ped_daemon runs in a window on phucsia (which has been ssh'ed to phast).  The output looks similar to the daq_daemon, except it will occasionally run pedestals.

If it crashes, CALL BURT.  

The deletor_daemon

deletor_daemon runs as as independent daemon on vmesparc.  It does not print out visible output, but logs itself to /var/log/deletor.log on vmesparc.

The pedestal selector

The pedestal selector runs on phucsia and looks like this: 

If it has crashed, CALL BURT.  (do you sense a theme here?)

"Why isn't the datamover sinking my sequence / running peds?"

Cause #1: the DAQ is currently writing to the partition which contains the next sequence scheduled for pedestal running. You can verify this via the run log, which lists the partition N in the filename (/data/N/PhoRaw...).

Solution: if you're taking ALLBEAM data, wait until the DAQ starts a new sequence. If you're taking PADDLECOSM or junk, stop the run and start a new run (PADDLECOSM @ 5k/event @ 20 Hz takes
more than 2 hours to fill up one sequence ...)

Cause #2: The pedestals or gains need to be manually validated.  The SiCalib alarm went off and was ignored in the heat of the moment.  Check the ped_daemon window -- if peds need to be validated, it will ask you to press ENTER to begin validation.  Consult with your local Si expert.