Hi Tj, many thanks for the pointer!Round robin is false so that it should be just the configured route as hard paths.
The router logs are a bit odd - AFAIS the routes are near their their limits. There were cycles on the CE where all jobs got send over the default "Condor_Pool" route [1] - but then with the next cycle the jobs then go via the "DESYNAF" route again [2]. I.e., the two cycles [1] and [2] 10secs apart.
For the moment, I have dropped the `Condor_Pool` default route to force everything on the `DESYNAF` route, but tbh I have no idea, why route decisions like [1] happened occasionally ð
Cheers, Thomas [1] 12/03/25 16:15:45 === Current Probing Information === 12/03/25 16:15:45 fsize: 45718426 mtime: 1764774943 12/03/25 16:15:45 first log entry: 102 CreationTimestamp 1756886501 12/03/25 16:15:45 JobRouter: Checking for candidate jobs. routing table is:Route Name Source Submitted/Max Idle/Max Throttle Recent: Started Succeeded Failed DESYNAF DESYNAF 94/ 10000 35/ 2000 10.2976 jobs/sec 4 2 0 Condor_Pool Condor_Pool 26/ 10000 9/ 2000 0.312857 jobs/sec 0 0 4 12/03/25 16:15:45 JobRouter: Umbrella constraint: (target.JobUniverse =?= 5 || target.JobUniverse =?= 1) && ( (True) || (Owner == "atlasnf1") ) && (target.ProcId >= 0 && target.JobStatus == 1 && (target.StageInStart is undefined || target.StageInFinish isnt undefined) && target.Managed isnt "ScheddDone" && target.Managed isnt "External" && target.Owner isnt Undefined && target.RoutedBy isnt "htcondor-ce") 12/03/25 16:15:45 JobRouter (src=114271.0,dest=164743.0): job mirror synchronized; removing job from internal 'retirement' status 12/03/25 16:15:45 JobRouter (src=114226.0,dest=164755.0): job mirror synchronized; removing job from internal 'retirement' status 12/03/25 16:15:45 JobRouter (src=114313.0,dest=164725.0): job mirror synchronized; removing job from internal 'retirement' status 12/03/25 16:15:45 JobRouter (src=114272.0,dest=164756.0): job mirror synchronized; removing job from internal 'retirement' status 12/03/25 16:15:45 SharedPortClient: sent connection request to local schedd for shared port id schedd_993_b1a2
12/03/25 16:15:45 Setting CpusUsage = 0.0 12/03/25 16:15:45 Setting DiskUsage = 2750 12/03/25 16:15:45 Setting ImageSize = 3500 12/03/25 16:15:45 Setting MemoryUsage = ((ResidentSetSize + 1023) / 1024) 12/03/25 16:15:45 Setting NumJobStarts = 1 12/03/25 16:15:45 Setting RemoteSysCpu = 6.0 12/03/25 16:15:45 Setting RemoteUserCpu = 2.0 12/03/25 16:15:45 Setting ResidentSetSize = 3500 12/03/25 16:15:45 Setting ScratchDirFileCount = 33512/03/25 16:15:45 JobRouter (src=114306.0,dest=164726.0,route=Condor_Pool): updated job status 12/03/25 16:15:45 SharedPortClient: sent connection request to local schedd for shared port id schedd_993_b1a2
12/03/25 16:15:45 Setting CpusUsage = 0.0 ... ### default Condor_Pool route taken from here until next cycle [2] 12/03/25 16:15:55 === Current Probing Information === 12/03/25 16:15:55 fsize: 45727554 mtime: 1764774955 12/03/25 16:15:55 first log entry: 102 CreationTimestamp 1756886501 12/03/25 16:15:55 JobRouter: Checking for candidate jobs. routing table is:Route Name Source Submitted/Max Idle/Max Throttle Recent: Started Succeeded Failed DESYNAF DESYNAF 94/ 10000 35/ 2000 10.2976 jobs/sec 4 2 0 Condor_Pool Condor_Pool 26/ 10000 9/ 2000 0.312857 jobs/sec 0 0 4 12/03/25 16:15:55 JobRouter: Umbrella constraint: (target.JobUniverse =?= 5 || target.JobUniverse =?= 1) && ( (True) || (Owner == "atlasnf1") ) && (target.ProcId >= 0 && t arget.JobStatus == 1 && (target.StageInStart is undefined || target.StageInFinish isnt undefined) && target.Managed isnt "ScheddDone" && target.Managed isnt "External" &&
target.Owner isnt Undefined && target.RoutedBy isnt "htcondor-ce") 12/03/25 16:15:55 JobRouter: Found candidate job src=114226.0,route=DESYNAF 12/03/25 16:15:55 JobRouter: Found candidate job src=114271.0,route=DESYNAF 12/03/25 16:15:55 JobRouter: Found candidate job src=114272.0,route=DESYNAF 12/03/25 16:15:55 JobRouter: Found candidate job src=114313.0,route=DESYNAF12/03/25 16:15:55 SharedPortClient: sent connection request to local schedd for shared port id schedd_993_b1a2
12/03/25 16:15:55 JobRouter (src=114271.0,route=DESYNAF): claimed job12/03/25 16:15:55 JobRouter post-route transform OnExitHold: does not match job. skippping it. 12/03/25 16:15:55 Will use TCP to update collector bird-htc-master21.desy.de <131.169.56.132:9618?alias=bird-htc-master21.desy.de> 12/03/25 16:15:55 Trying to query collector <131.169.56.132:9618?alias=bird-htc-master21.desy.de> 12/03/25 16:15:55 SharedPortClient: sent connection request to schedd naf-htc-preprod-ce01.desy.de for shared port id schedd_1391_b053 12/03/25 16:15:55 SharedPortClient: sent connection request to schedd naf-htc-preprod-ce01.desy.de for shared port id schedd_1391_b053 12/03/25 16:15:55 JobRouter (src=114271.0,dest=164836.0,route=DESYNAF): submitted job 12/03/25 16:15:55 JobRouter (src=114271.0,dest=164836.0,route=DESYNAF): submitted job has not yet appeared in job queue mirror or was removed (submitted 0 seconds ago) 12/03/25 16:15:55 SharedPortClient: sent connection request to local schedd for shared port id schedd_993_b1a2
12/03/25 16:15:55 JobRouter (src=114226.0,route=DESYNAF): claimed job12/03/25 16:15:55 JobRouter post-route transform OnExitHold: does not match job. skippping it. 12/03/25 16:15:55 Will use TCP to update collector bird-htc-master21.desy.de <131.169.56.132:9618?alias=bird-htc-master21.desy.de> 12/03/25 16:15:55 Trying to query collector <131.169.56.132:9618?alias=bird-htc-master21.desy.de> 12/03/25 16:15:55 SharedPortClient: sent connection request to schedd naf-htc-preprod-ce01.desy.de for shared port id schedd_1391_b053 12/03/25 16:15:55 SharedPortClient: sent connection request to schedd naf-htc-preprod-ce01.desy.de for shared port id schedd_1391_b053
... ### DESYNAF taken from here ### On 2025-12-04 00:09, John M Knoeller via HTCondor-users wrote:
This sounds like   JOB_ROUTER_ROUND_ROBIN_SELECTION = true is configured in the CE. ÂThe default is false.   condor_ce_config_val -v JOB_ROUTER_ROUND_ROBIN_SELECTION Will show if that is the case.Alternatively, there is a limit on the total number of jobs that can be routed to DESYNAF and that limitis being hit.   condor_ce_job_router_info -config should show any limits on the routes. -tj ------------------------------------------------------------------------*From:*ÂHTCondor-users <htcondor-users-bounces@xxxxxxxxxxx> on behalf of Thomas Hartmann <thomas.hartmann@xxxxxxx>*Sent:*ÂWednesday, December 3, 2025 9:26 AM *To:*ÂHTCondor-Users Mail List <htcondor-users@xxxxxxxxxxx>*Subject:*Â[HTCondor-users] CondorCE: jobs taking different routes albeit the same route condition are fulfilledHi all, I am sometimes observing, that jobs on a test CE are taking different routes although the conditions should be the same for taking a specific route. I.e., some jobs go via the `DESYNAF` as intended but some go via the default `Condor_Pool` route [1] The two example CE jobs in [1] have both the same `atlasnf1` owner [2] mapped - which is actually the sole requirement for the `DESYNAF` route [3], so that I would expect both jobs to match to this route, or? Cheers,  Thomas [1] 12/03/25 16:05:53 JobRouter: Found candidate job src=114376.0,route=DESYNAF vs 12/03/25 15:38:32 JobRouter: Found candidate job src=114316.0,route=Condor_Pool [2] [root@naf-htc-preprod-ce01 config.d]# sdiff /tmp/114376.DESYNAF.ads /tmp/114316.CondorPool.ads | grep -i owner Owner = "atlasnf1" Owner = "atlasnf1" [3] [root@naf-htc-preprod-ce01 config.d]# cat 99_nafroute.conf JOB_ROUTER_ROUTE_DESYNAF @=end  UNIVERSE VANILLA  REQUIREMENTS Owner == "atlasnf1"  ... @end _______________________________________________ HTCondor-users mailing list To unsubscribe, send a message to htcondor-users-request@xxxxxxxxxxx with a subject: Unsubscribe The archives can be found at: https://www-auth.cs.wisc.edu/lists/htcondor-users/
Attachment:
smime.p7s
Description: S/MIME Cryptographic Signature