making it possible to claim relative hosts from the hostfile/scheduler
by using +n# hostname, where 0 <= # < np
ex:
cat ~/work/svn/hpc/dev/test/Rankfile/rankfile
rank 0=+n0 slot=0
rank 1=+n0 slot=1
rank 2=+n1 slot=2
rank 3=+n1 slot=1
This commit was SVN r21557.
prepare the send buffer, and post the collective order to the local daemon. It
then register the callback and return fromthe modex exchange. It will only
wait for this modex completion when the modex_recv is called. Meanwhile, the
daemon will do the allgather.
This commit was SVN r21543.
Add two options for plm process module, i.e. remote_env_prefix for getting OMPI prefix on remote computer by reading its user environment variable (OPENMPI_HOME), and remote_reg_prefix is similar, but it reads the registry on the remote computer. Reading remote env prefix has a higher priority than reading reg prefix, so that user can use their own installation of OMPI.
This commit was SVN r21538.
It never set the daemon launch counter before launching.
In the output for 'total job launch time' make the message match that of the SLURM PLM for easier parsing.
This commit was SVN r21527.
coded binomial approach. Now, the PLM extract the children information from the
routed component, and the startup follow the routed topology. As a side effect,
we can now launch using linear or radix topologies, in addition to the previously
binomial topology.
This commit was SVN r21514.
the message to update the structures, but instead use the information from
the URI. The reason is that even the launch report messages can get routed.
Deal with the orted_cmd_line in a single location.
This commit was SVN r21513.
Normally, any non-windows nodes should return NodeStatus_Unreachable, but that's not true. We have noticed that Scientific Linux cluster nodes, which are in the same subnet as the Windows nodes, return NodeStatus_PendingApproval value, and the RAS CCP takes them as Windows nodes. So let's just use the "Ready" nodes.
This commit was SVN r21510.
that now the HNP send the messages using the routed component. In the case
of tree spawn, when a intermediary node spawn a child it doesn't know how
to forward a message to it, so when the node-map message is coming from
the HNP (as there is nothing yet in the contact/routing table) the message
is sent back the way it came. As a result the node-map message keeps jumping
between the HNP and the first level orteds.
The solution is to add a new option to the children orte_parent_uri, which
is only set when the orted is _not_ directly spawned by the HNP. When this
option is present on the argument list, the orted will add the parent to
its routing, and force the parent to update his routes (by sending the URI).
With this approach, the routing tree is build in same time as the processes
are spawned, and all messages from the HNP can be routed to the leaves.
However, this is far from an optimal solution. Right now, this so called tree
spawn, only spawn the children in a tree without doing anything about the
"connect back to the HNP" step. The HNP is flooded with reports from all the
orted. The total number of messages is higher than in the non tree startup
scheme, so we do not expect this approach to be scalable in the current
incarnation. A complete overhaul of the tree startup is required in order
improve the scalability. Stay tuned!
This commit was SVN r21504.
The open/select of the PLM is done in orte/mca/ess/base/ess_base_std_orted.c. It only is done when the PLM MCA param is set directing a specific PLM be selected. The function
orte_plm_base_orted_append_basic_args
clears the params passed to the daemon of any PLM selection passed to the HNP. Each PLM then adds a PLM directive if-and-only-if backend PLM support is desired. At present, Torque, SLURM, and rsh all specify this support and direct that the backend orted open the "rsh" PLM.
This commit was SVN r21488.
The following SVN revision numbers were found above:
r21480 --> open-mpi/ompi@ed585bce8a
If we do not initialize the PML, non-HNP daemons will not be able to use its functions. For example, RSH needs it when the tree_spawn mode is
enabled: daemons call orte_pml.remote_spawn() function to spawn their children in the deployment tree.
This commit was SVN r21480.
