Itâs more like there shall only be one condor_negotiator unless you shard your pool, and it is most efficient to put it on same machine as the condor_collector. It would almost certainly be a mistake to have a NEGOTIATOR in the daemon list of an execute node.
A quick way to check would be to run
condor_status -master -constraint âNEGOTIATOR_Timestampâ -af:h Name âformattime(NEGOTIATOR_Timestamp)â
From the Administratorsâ Manual Introduction I can conclude (frankly without confidence) the condor_negotiator shall run on the Central Manager node, but I see condor_negotiator output (in fresh rsyslog output setup) also on the Execute and Submit nodes. Is this OK, or there is a mistake in our HT Condor configuration?