[R-sig-hpc] Problem Running 'snow' on Several Machines

Michael Smith my.r.help at gmail.com
Thu Apr 28 07:22:51 CEST 2011


Steve,

Thank you so much! When I'm on 192.168.1.100 and I call it like

    > cl <- makeSOCKcluster(c("192.168.1.100", "192.168.1.101"),
master="192.168.1.100")

everything works. The reason is that somehow Fedora does not "broadcast"
the hostname on DHCP.

Best,
Michael


On 04/28/2011 12:54 PM, Stephen Weston wrote:
> Unfortunately, there are dozens of possible reasons that makeSOCKcluster
> could hang, so it's impossible for us to guess what your particular problem is.
> You may have several issues to resolve before you'll get this running if
> you're unlucky.
> 
> However, one common problem is that one of the workers (in your case
> "192.168.1.101") may be unable to connect back to the master process using
> the default name used by makeSOCKcluster.  It may work better to use the
> dot-separated IP address, which you might be able to compute using:
> 
>     > master <- nsl(Sys.info()["nodename"])
> 
> Then create the cluster with:
> 
>     > cl <- makeSOCKcluster(c("localhost", "192.168.1.101"), master=master)
> 
> But if that shot in the dark doesn't work, I suggest debugging it by setting
> the manual argument to TRUE:
> 
>     > cl <- makeSOCKcluster(c("localhost", "192.168.1.101"), manual=TRUE)
> 
> It will then tell you the command to use to start each of the workers.
> Open a terminal on each worker, and execute the commands as it prompts you.
> Now, you should see the error message, and finding the solution will be
> much easier.
> 
> - Steve
> 
> 
> On Fri, Apr 22, 2011 at 5:30 AM, Michael Smith <my.r.help at gmail.com> wrote:
>> All,
>>
>> I want to run the following program:
>>
>> library("snow")
>> cl <- makeSOCKcluster(c("localhost", "192.168.1.101"))
>> clusterCall(cl, function() Sys.info()[c("nodename","machine")])
>> stopCluster(cl)
>>
>> I have set up SSH keys so I don't need to type a password when logging
>> in to 192.168.1.101. I'm using Fedora 13 on both machines.
>>
>> What happens is that the program somehow hangs after 'makeSOCKcluster'
>> because it does not return to the R prompt and I have to hit Ctrl-C. It
>> also fails to establish the cluster:
>>
>> $ R -q
>>> library("snow")
>>
>> Attaching package: 'snow'
>>
>> The following object(s) are masked from 'package:base':
>>
>>    enquote
>>
>>> cl <- makeSOCKcluster(c("localhost", "192.168.1.101"))
>>
>> Attaching package: 'snow'
>>
>> The following object(s) are masked from 'package:base':
>>
>>    enquote
>>
>>
>> Attaching package: 'snow'
>>
>> The following object(s) are masked from 'package:base':
>>
>>    enquote
>>
>> ^C
>>> clusterCall(cl, function() Sys.info()[c("nodename","machine")])
>> Error in inherits(cl, "cluster") : object 'cl' not found
>>
>>
>> When I use makeSOCKcluster(c("localhost", "localhost")) everything works
>> fine locally.Below is more information on the system.
>>
>> Thanks,
>> Michael
>>
>>
>>
>>> sessionInfo()
>> R version 2.12.2 (2011-02-25)
>> Platform: x86_64-redhat-linux-gnu (64-bit)
>>
>> locale:
>>  [1] LC_CTYPE=en_US.utf8       LC_NUMERIC=C
>>  [3] LC_TIME=en_US.utf8        LC_COLLATE=en_US.utf8
>>  [5] LC_MONETARY=C             LC_MESSAGES=en_US.utf8
>>  [7] LC_PAPER=en_US.utf8       LC_NAME=C
>>  [9] LC_ADDRESS=C              LC_TELEPHONE=C
>> [11] LC_MEASUREMENT=en_US.utf8 LC_IDENTIFICATION=C
>>
>> attached base packages:
>> [1] stats     graphics  grDevices utils     datasets  methods   base
>>
>> other attached packages:
>> [1] snow_0.3-3
>>> Sys.getlocale()
>> [1]
>> "LC_CTYPE=en_US.utf8;LC_NUMERIC=C;LC_TIME=en_US.utf8;LC_COLLATE=en_US.utf8;LC_MONETARY=C;LC_MESSAGES=en_US.utf8;LC_PAPER=en_US.utf8;LC_NAME=C;LC_ADDRESS=C;LC_TELEPHONE=C;LC_MEASUREMENT=en_US.utf8;LC_IDENTIFICATION=C"
>>> library(help=snow)
>>
>>
>>
>>
>>                Information on package 'snow'
>>
>> Description:
>>
>> Package:       snow
>> Title:         Simple Network of Workstations
>> Version:       0.3-3
>> Author:        Luke Tierney, A. J. Rossini, Na Li, H. Sevcikova
>> Description:   Support for simple parallel computing in R.
>> Maintainer:    Luke Tierney <luke at stat.uiowa.edu>
>> Suggests:      Rmpi,rpvm,rlecuyer,rsprng,nws
>> License:       GPL
>> Depends:       R (>= 2.3), utils
>> Packaged:      Fri Jul 4 18:07:30 2008; luke
>> Built:         R 2.12.2; ; 2011-04-21 03:02:00 UTC; unix
>>
>> Index:
>>
>> clusterSetupRNG         Uniform Random Number Generation in SNOW
>>                        Clusters
>> clusterSplit            Cluster-Level SNOW Functions
>> getMPIcluster           Starting and Stopping SNOW Clusters
>> parLapply               Higher Level SNOW Functions
>>
>> _______________________________________________
>> R-sig-hpc mailing list
>> R-sig-hpc at r-project.org
>> https://stat.ethz.ch/mailman/listinfo/r-sig-hpc
>>



More information about the R-sig-hpc mailing list