Merge pull request #74 from bjarthur/bja/lsf

add support for LSF

README.md

@@ -6,7 +6,9 @@ Support for different job queue systems commonly used on compute clusters.
 
 | Job queue system | Command to add processors |
 | ---------------- | ------------------------- |
+| Load Sharing Facility (LSF) | `addprocs_lsf(np::Integer, flags=``)` or `addprocs(LSFManager(np, flags))` |
 | Sun Grid Engine  | `addprocs_sge(np::Integer, queue="")` or `addprocs(SGEManager(np, queue))` |
+| SGE via qrsh | `addprocs_qrsh(np::Integer, queue="")` or `addprocs(QRSHManager(np, queue))` |
 | PBS              | `addprocs_pbs(np::Integer, queue="")` or `addprocs(PBSManager(np, queue))` |
 | Scyld | `addprocs_scyld(np::Integer)` or `addprocs(ScyldManager(np))` |
 | HTCondor | `addprocs_htc(np::Integer)` or `addprocs(HTCManager(np))` |
@@ -98,6 +100,20 @@ julia>  From worker 26: lum-7-2.local
         From worker 25: cheech-207-16.local
 ```
 
+### SGE via qrsh
+
+`SGEManager` uses SGE's `qsub` command to launch workers, which communicate the
+TCP/IP host:port info back to the master via the filesystem.  On filesystems
+that are tuned to make heavy use of caching to increase throughput, launching
+Julia workers can frequently timeout waiting for the standard output files to appear.
+In this case, it's better to use the `QRSHManager`, which uses SGE's `qrsh`
+command to bypass the filesystem and captures STDOUT directly.
+
+### Load Sharing Facility (LSF)
+
+`LSFManager` supports IBM's scheduler.  Similar to `QRSHManager` in that it
+uses the `-I` (i.e. interactive) flag to `bsub`.
+
 ### Using `LocalAffinityManager` (for pinning local workers to specific cores)
 
 - Linux only feature.

src/ClusterManagers.jl

@@ -17,5 +17,6 @@ include("condor.jl")
 include("slurm.jl")
 include("affinity.jl")
 include("elastic.jl")
+include("lsf.jl")
 
 end

src/lsf.jl

@@ -0,0 +1,41 @@
+export LSFManager, addprocs_lsf
+
+struct LSFManager <: ClusterManager
+    np::Integer
+    bsub_flags::Cmd
+end
+
+function launch(manager::LSFManager, params::Dict, launched::Array, c::Condition)
+    try
+        dir = params[:dir]
+        exename = params[:exename]
+        exeflags = params[:exeflags]
+
+        np = manager.np
+
+        jobname = `julia-$(getpid())`
+
+        cmd = `$exename $exeflags $(worker_arg())`
+        bsub_cmd = `bsub -I $(manager.bsub_flags) -cwd $dir -J $jobname "$cmd"`
+
+        stream_proc = [open(bsub_cmd) for i in 1:np]
+
+        for i in 1:np
+            config = WorkerConfig()
+            config.io = stream_proc[i]
+            push!(launched, config)
+            notify(c)
+        end
+
+    catch e
+        println("Error launching workers")
+        println(e)
+    end
+end
+
+manage(manager::LSFManager, id::Int64, config::WorkerConfig, op::Symbol) = nothing
+
+kill(manager::LSFManager, id::Int64, config::WorkerConfig) = kill(config.io)
+
+addprocs_lsf(np::Integer, bsub_flags::Cmd=``; params...) =
+        addprocs(LSFManager(np, bsub_flags); params...)