ClusterWare 6/7 vs. Current ClusterWare

ClusterWare 6/7	bpsh 0-127 date
ClusterWare	pdsh -w n[0-127] date
ClusterWare	scyld-nodectl -i n[0-127] exec date

Attempt to remotely execute the date command on all nodes in the range of n0 to n127.

bpsh executes on the master node or compute nodes. Authentication is set up out-of-the-box for user root, though it must be set up by the local cluster administrator for non-root users.

pdsh executes on the head node or compute nodes and requires proper user authentication between nodes.

scyld-nodectl executes on any node that has the clusterware-tools package installed and is only available to cluster administrators.

ClusterWare 6/7	bpsh -a date
ClusterWare	scyld-nodectl --up exec date

Perform the same remote execution of the date command, although this time just for nodes in the "up" state.

ClusterWare 6/7	bpcp
ClusterWare	rsync, scp, pdcp
ClusterWare	scyld-nodectl -i <nodes> scp

bpcp executes on the master node or compute nodes. Authentication is set up out-of-the-box for user root, though it must be set up by the local cluster administrator for non-root users.

rsync, scp, pdcp execute on the head node or compute nodes, and each require proper user authentication between nodes.

ClusterWare 6/7	bpstat
ClusterWare	scyld-nodectl status

ClusterWare 6/7	bpstat -U
ClusterWare	scyld-nodectl status --refresh

ClusterWare 6/7	bpstat --long
ClusterWare	scyld-nodectl status --long

bpstat only executes on the master node.

scyld-nodectl executes on any node that has the clusterware-tools package installed and is only available to cluster administrators.

ClusterWare 6/7	beostat, beostatus
ClusterWare	Monitoring GUI

See monitoring_gui for a brief discussion of graphical monitoring and a further reference to full documentation, and monitoring_scyld-nodectl for examples of the command-line interface for accessing partial per-node information.

ClusterWare 6/7	beomap
ClusterWare	unavailable

Legacy ClusterWare anticipated users executing multi-threaded jobs by using beomap and beostat to determine the availability and current load-levels of the compute nodes, and then executing MPI-type commands (e.g., mpirun) from the head node to execute a job on selected nodes. Modern ClusterWare users instead rely upon a job scheduler (e.g., Slurm or PBS TORQUE) to optimally manage executing such multi-threaded jobs across multiple nodes.

ClusterWare 6/7	bpctl -S 0-127 -R
ClusterWare	scyld-nodectl -i n[0-127] reboot

ClusterWare 6/7	bpctl -S all -R
ClusterWare	scyld-nodectl -a reboot

ClusterWare 6/7	bpctl -S allup -R
ClusterWare	scyld-nodectl --up reboot

bpctl only executes on the master node.

scyld-nodectl executes on any node that has the clusterware-tools package installed and is only available to cluster administrators.

ClusterWare 6/7	beoconfig node
ClusterWare	scyld-nodectl ls -l

ClusterWare 6/7	beoconfig kernelimage
ClusterWare	scyld-bootctl -a ls -l

ClusterWare 6/7	beoconfig kernelcommandline
ClusterWare	scyld-imgctl -a ls -l

beoconfig only executes on the master node.

scyld-nodectl, scyld-bootctl, scyld-imgctl execute on any node that has the clusterware-tools package installed and is only available to cluster administrators.