mismatch betwen information of partition nodes at compendium and real slurm cfg.

At the following pages the information about which nodes belongs to which partition doesn't match the actual configuration of the system.

https://doc.zih.tu-dresden.de/jobs_and_resources/hardware_overview/#island-4-to-6-intel-haswell-cpus

https://doc.zih.tu-dresden.de/jobs_and_resources/partitions_and_limits/#memory-limits

For instance, nodes: taurusi[6541-6558] are at haswell256, not at haswell64

added Content New label

Consider to create and deploy an process building the partition and limits table

WIP, It shows the partition name and the amount of nodes only.

Example:

sh bin/gentable.sh alpha
query: alpha
------------------------
Partition: alpha
Nodes: taurusi[8003-8034]
Total nodes in partition: 32
----
Partition: alpha-interactive
Nodes: taurusi[8001-8002]
Total nodes in partition: 2
----

#!/bin/bash

function show_nodes(){

  for line in ${PARTITIONS[@]}; do
    [ ! -z $DEBUG ] && echo "$line"
    PARTITION_NAME=$(echo $line | awk '{print $1}' )
    NODES=$(echo $line | awk '{print $2}' )
    NNODES=$(echo $line | awk '{print $3}' )
    NCPUS=$(echo $line | awk '{print $4}' )
    echo "Partition: $PARTITION_NAME"
    echo "Nodes: $NODES"
    echo "Total nodes: $NNODES - Total CPUs: $NCPUS"
  done
}

function s_info(){
  # get partition name, total nodes, totalCPUs and node list 
  AWK_CMD='/^PartitionName='$QUERY'/ {print $1"\n "$28" "$27"\n "$18"\n"}'

  if [ ! -z $file_name ]; then
    IFS=$'^\n' \
    PARTITIONS=( $(awk "$AWK_CMD" RS= $file_name | awk -F"=| |\n" '{print $2" "$10" "$5" "$7}' RS=) )
  else
    IFS=$'^\n' \
    PARTITIONS=( $(scontrol show part | awk "$AWK_CMD" RS= | awk -F"=| |\n" '{print $2" "$10" "$5" "$7}' RS=) )
  fi

  for line in ${PARTITIONS[@]}; do
    PARTITION_NAME=$(echo $line | awk '{print $1}' )
    sinfo -p $PARTITION_NAME -h -o "%R %N %D %c %e %m %G"
  #  CPU_PER_NODE=$(sinfo -p $PARTITION_NAME --noheader -o %c)
  
  done
}

function count_nodes(){

  # get partition name, total nodes, totalCPUs and node list 
  AWK_CMD='/^PartitionName='$QUERY'/ {print $1"\n "$28" "$27"\n "$18"\n"}'

  if [ ! -z $file_name ]; then
    IFS=$'^\n' \
    PARTITIONS=( $(awk "$AWK_CMD" RS= $file_name | awk -F"=| |\n" '{print $2" "$10" "$5" "$7}' RS=) )
  else
    IFS=$'^\n' \
    PARTITIONS=( $(scontrol show part | awk "$AWK_CMD" RS= | awk -F"=| |\n" '{print $2" "$10" "$5" "$7}' RS=) )
  fi


for line in ${PARTITIONS[@]}; do
  PARTITION_NAME=$(echo $line | awk '{print $1}' )
  NODES=$(echo $line | awk '{print $2}' )
  NNODES=$(echo $line | awk '{print $3}' )
  NCPUS=$(echo $line | awk '{print $4}' )
  CPU_PER_NODE=$(sinfo -p $PARTITION_NAME --noheader -o %c)
  echo "Partition: $PARTITION_NAME"
  echo "Nodes: $NODES"

  # get nodes list in between commas
  IFS=$' ' \
  NODE_LIST="$(echo $NODES | sed 's/,\([^{0-9}]\)/ \1/g')"

  NODE_LIST=( $NODE_LIST )

  if [ ! -z $verbose ]; then
    echo " node list size: ${#NODE_LIST[@]}"
  fi

  nnodes_part=0
  for node in ${NODE_LIST[@]}; do
    node_name=$(echo $node | sed 's/\[.*\]//' )

    node_range="$(echo $node | grep -o "[0-9,-]*" | sed 's/,/ /g')" # this doesn't work well cos get's nodes names with number 
    #
    node_range="$(echo $node | grep -o "\[[0-9,-]*\]" | sed -e 's/\[\|\]//g' -e 's/,/ /g')"
    node_range=( $node_range )
    if [ ! -z $verbose ]; then
      echo " - nodes: $node"
      echo " - node name: $node_name range: ${node_range[@]} - ${#node_range[@]}"
    fi

    total_nodes=1
    if [ ${#node_range[@]} -gt 0 ]; then
      total_nodes=0
      # count the amount of nodes in the ranges
      for range in ${node_range[@]}; do
        sum_nodes=1
        range_0=$(echo $range | awk -F- '{print $1}')
        range_n=$(echo $range | awk -F- '{print $2}')
        sum_nodes=$(( sum_nodes + ( $range_n - $range_0 )))
        if [ ! -z $DEBUG ]; then
          echo "   - $range = $sum_nodes"
        fi
        total_nodes=$(( total_nodes + sum_nodes ))
      done
      node_test="$node_name$range_0"
    else
      node_test="$node_name"
    fi

    if [ ! -z $verbose ]; then
      echo "  Total nodes: $total_nodes"
    fi

    nnodes_part=$(( $nnodes_part + total_nodes ))

  done
  echo "Total nodes in partition: $nnodes_part"
  #sinfo -p $PARTITION_NAME --noheader -O Partition,CPUs,MaxCPUsPerNode,Memory,Gres
  echo ---- 
done
}

help()
{
   echo "get slurm partition information."
   echo
   echo "Syntax: $(basename $0) [-h|-f|v] [partition-name]"
   echo "options:"
   echo "-h         Print this Help."
   echo "-v         be verbose."
   echo "-f         use file <file_name>"
   echo
}


while getopts "hvf:" option; do
   case $option in
      h)
         help
         exit 0
         ;;
      f)
         file_name=$OPTARG
         ;;
      v)
         echo "verbose"
         verbose=1
         ;;
     \?)
         echo "Error: Invalid option"
         exit 1
         ;;
   esac
done
shift $((OPTIND-1))

if [ "$#" -gt "1" ]; then
  echo "Error: query should be one string"
  exit 1
fi

QUERY=$@

if [ ! -z "$QUERY" ]; then
  echo "query: $QUERY"
fi

if [ ! -z $file_name ]; then
  echo "using file: $file_name"
  if [ ! -f $file_name ]; then
    echo "no such file $file_name"
    exit 1
  fi
fi

echo "------------------------"
#show_nodes
count_nodes $QUERY
#s_info

mentioned in merge request !736 (merged)

Resolved as we now create partitions table from Slurm.

closed