[R] Rmpi with LAM/MPI on Debian

[Dirk Eddelbuettel]

On 6 February 2009 at 05:15, Ingeborg Schmidt wrote:
| Hello,
| I seem to be unable to get Lam/Mpi to work with Rmpi on Debian 5.

The Debian Rmpi package (ie r-cran-rmpi) switched to Open MPI instead of
LAM/MPI quite some time ago as LAM/MPI is no longer actively developed, and
the LAM/MPI team itself recommends switching to Open MPI. So recent versions
of r-cran-rmpi will not work with LAM/MPI.

| Following the readme file, I installed the packages r-cran-rmpi and lam3 lam3-dev lam-runtime . The lam-bhost.def is configured like this:

You can of course uninstall r-cran-rmpi, and then use the LAM/MPI development
package (currently called liblam4 on my Debian testing system) to locally
intall Rmpi into /usr/local/lib/R/site-packages via R CMD INSTALL.  You will
have have to follow the instruction in the Rmpi sources.
 
| hn01
| cn01 cpu=4
| cn02 cpu=4
| 
| Lam seems to be fine with this:
| 
| >lamboot -v
| n-1<13746> ssi:boot:base:linear: booting n0 (hn01)
| n-1<13746> ssi:boot:base:linear: booting n1 (cn01)
| n-1<13746> ssi:boot:base:linear: booting n2 (cn02)
| n-1<13746> ssi:boot:base:linear: finished
| 
| However, Rmpi 0.5.5 is not:
| 
|  > library("Rmpi")
| [hn01:30424] mca: base: component_find: unable to open osc pt2pt: file 

These are __Open MPI__ error messages.  You are mixing systems. 

There is a mailing list r-sig-hpc for these question. Please consider
subscribing and posting there.

Hth, Dirk

| not found (ignored)
|  > mpi.spawn.Rslaves()
|     1 slaves are spawned successfully. 0 failed.
| master (rank 0, comm 1) of size 2 is running on: hn01
| slave1 (rank 1, comm 1) of size 2 is running on: hn01
|  > mpi.remote.exec(paste("I am",mpi.comm.rank(),"of",mpi.comm.size()))
| $slave1
| [1] "I am 1 of 2"
|  > mpi.close.Rslaves()
| [1] 1
|  > mpi.quit()
| 
| Note how only one slave is spawned. Trials with mpiexec -n 3 R --no-save -q and the .Rprofile in place started R on multiple nodes, but produced this output:
| 
| [hn01:04612] mca: base: component_find: unable to open osc pt2pt: file not found (ignored)
| > [cn01:30306] mca: base: component_find: unable to open osc pt2pt: file not found (ignored)
| [cn01:30304] mca: base: component_find: unable to open osc pt2pt: file not found (ignored)
| [cn01:30305] mca: base: component_find: unable to open osc pt2pt: file not found (ignored)
| > 
| [1]> 
| [1]> 
| [1] "Please use mpi.quit() to quit R"
|  "Please use mpi.quit() to quit R"
|  "Please use mpi.quit() to quit R"
| -----------------------------------------------------------------------------
| It seems that [at least] one of the processes that was started with
| mpirun did not invoke MPI_INIT before quitting (it is possible that
| more than one process did not invoke MPI_INIT -- mpirun was only
| notified of the first one, which was on node n0).
| 
| mpirun can *only* be used with MPI programs (i.e., programs that
| invoke MPI_INIT and MPI_FINALIZE).  You can use the "lamexec" program
| to run non-MPI programs over the lambooted nodes.
| -----------------------------------------------------------------------------
| mpirun failed with exit status 252
| 
| Do you have any suggestions?
| 
| Cheers,
| Ingeborg Schmidt
| 
| 
| 
| 
| 
| ______________________________________________
| R-help at r-project.org mailing list
| https://stat.ethz.ch/mailman/listinfo/r-help
| PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
| and provide commented, minimal, self-contained, reproducible code.

-- 
Three out of two people have difficulties with fractions.