Resurfacing of Bug #16213 : problem with two eventbuilders sending to two dataloggers
The issue described in Bug #16213 (https://cdcvs.fnal.gov/redmine/issues/16213), where events stopped flowing after a few seconds (at 10 events/s) with two eventbuilders sending events to two dataloggers, appears to have resurfaced. Essentially, having installed artdaq-demo on woof via "./quick-mrb-start.sh --tag=v2_10_00", I have no problem running in a 2x2x2 scenario where each datalogger receives events from just one of the two eventbuilders, but I get a timeout failure if I try having both eventbuilders send to both data loggers. The FHiCL documents for the former case can currently be found in woof:/home/jcfree/scratch/artdaq-demo_v2_10_00/fhicl/each_eventbuilder_to_specific_datalogger , and for the latter case in woof:/home/jcfree/scratch/artdaq-demo_v2_10_00/fhicl/both_eventbuilders_to_both_dataloggers . You can also see this phenomenon if you look at info for one of the runs I took with this artdaq-demo installation, woof:/home/jcfree/scratch/artdaq-demo_v2_10_00/run_records/8 .
#1 Updated by Eric Flumerfelt almost 3 years ago
Your FHiCL documents are missing the "daq.aggregator.auto_suppression_enabled: false" parameter that tells the DataReceiverManager not to suppress sources based on uneven receive counts. Since this is a symmetric system, DL1 will only receive events from EB1 and DL2 will only receive events from EB2. A check on this being the problem would be to run a 2x3x2 system, this should not have this issue (as each DataLogger would then be receiving events from all EventBuilders).
#2 Updated by Eric Flumerfelt almost 3 years ago
- Project changed from artdaq to artdaq Utilities
- Category set to artdaq-daqinterface
- Priority changed from High to Normal
I'm moving this issue over to artdaq_daqinterface, as it needs to make this change. artdaq-demo/tools/generateAggregator.rb already puts the auto_suppression_enabled:false into the Aggregator code, for example.