Task #10059: Merge cobalt partition into cpu partition, and strictly use 2 cores/subband

CEP/Pipeline/recipes/sip/tasks.cfg.CEP4.in

@@ -15,12 +15,12 @@ nproc = 0
 nproc = 0
 
 [imager_prepare]
-nthreads = 10
+nthreads = 2
 
 [long_baseline]
 nproc = 0
 rficonsole_executable = /opt/aoflagger/bin/aoflagger
-nthreads = 10
+nthreads = 2
 
 [dppp]
 max_per_node = 0 
@@ -28,34 +28,34 @@ nthreads = 2
 
 [awimager]
 max_per_node = 0 
-nthreads = 10
+nthreads = 2
 
 [rficonsole]
 executable = /opt/aoflagger/bin/aoflagger
 max_per_node = 0 
-nthreads = 10
+nthreads = 2
 
 [imager_prepare]
 rficonsole_executable = /opt/aoflagger/bin/aoflagger
-nthreads = 10
+nthreads = 2
 
 [imager_bbs]
-nthreads = 10
+nthreads = 2
 
 [bbs_reducer]
-nthreads = 10
+nthreads = 2
 
 [executable_args]
-nthreads = 10
+nthreads = 2
 
 [casapy-imager]
-nthreads = 10
+nthreads = 2
 
 [pythonplugin]
-nthreads = 10
+nthreads = 2
 
 [python-calibrate-stand-alone]
-nthreads = 10
+nthreads = 2
 
 [calibrate-stand-alone]
-nthreads = 10
+nthreads = 2

MAC/Services/src/PipelineControl.py

@@ -145,36 +145,25 @@ class Parset(dict):
 
   def processingNumberOfCoresPerTask(self):
     result = int(self[PARSET_PREFIX + "Observation.Cluster.ProcessingCluster.numberOfCoresPerTask"]) or "21"
-    if result != 21:
+    if result != 2:
         logger.warn('Invalid Observation.Cluster.ProcessingCluster.numberOfCoresPerTask: %s, defaulting to %s', result, 21)
-    return 21
+    return 2
 
   def processingNumberOfTasks(self):
     """ Parse the number of nodes to allocate from "Observation.Cluster.ProcessingCluster.numberOfTasks",
         which can have either the format "{number}" or "{min}-{max}". """
 
-    defaultValue = "12-25"
+    defaultValue = 244
     parsetValue = self[PARSET_PREFIX + "Observation.Cluster.ProcessingCluster.numberOfTasks"].strip()
 
-    if "-" in parsetValue:
-      # min,max
-      _min, _max = parsetValue.split("-")
-
-      # collapse if not min <= max
-      if _min > _max:
-        result = _min
-      else:
-        result = "%s-%s" % (_min, _max)
-    else:
-      # plain number
-      result = int(parsetValue)
+    result = int(parsetValue) * 10 # ScS expects to schedule 10 jobs/node, and still specifies #nodes
 
     # apply bound
-      if result < 1 or result > 50:
+    if result <= 0 or result > 50*24:
       result = defaultValue
 
     if result != parsetValue:
-      logger.error('Invalid Observation.Cluster.ProcessingCluster.numberOfTasks: %s, defaulting to %s', parsetValue, result)
+      logger.warn('Invalid Observation.Cluster.ProcessingCluster.numberOfTasks: %s, defaulting to %s', parsetValue, result)
 
     return result
 
@@ -437,7 +426,7 @@ class PipelineControl(OTDBBusListener):
                      "--nice=1000",
 
                      "--partition=%s" % parset.processingPartition(),
-                     "--nodes=%s" % parset.processingNumberOfTasks(),
+                     "--ntasks=%s" % parset.processingNumberOfTasks(),
                      "--cpus-per-task=%s" % parset.processingNumberOfCoresPerTask(),
 
                      # Define better places to write the output

RTCP/Cobalt/GPUProc/src/scripts/cobalt_functions.sh

@@ -52,7 +52,7 @@ function read_cluster_model {
   case "${CLUSTER_NAME}" in
     CEP4)
       HEADNODE=head01.cep4.control.lofar
-      COMPUTENODES="`ssh $HEADNODE sinfo --responding --states=idle,mixed,alloc --format=%n.cep4,%T --noheader --partition=cobalt --sort=N | fgrep -v ,draining | cut -f1 -d,`"
+      COMPUTENODES="`ssh $HEADNODE sinfo --responding --states=idle,mixed,alloc --format=%n.cep4,%T --noheader --partition=cpu --sort=N | fgrep -v ,draining | cut -f1 -d,`"
       if [ -z "$COMPUTENODES" ]; then
         echo "ERROR: Could not obtain list of available CEP4 nodes. Defaulting to all."
         COMPUTENODES="`seq -f "cpu%02.0f.cep4" 1 50`"