[R-sig-hpc] snow job

[Stephen Weston]

Using "manual = TRUE" is particularly useful when the worker
gets an error before creating the log file.  And using it doesn't take
any R skills at all, although I guess you do need to know how to
type a command into a command window.  I suggest you learn
how to use manual mode, because you will probably need it for
debugging problems in the future.

You just need to open a new command window, and paste or
type the command that makeSOCKcluster tells you to execute.
Hopefully you'll see an error message.

- Steve


On Wed, May 11, 2011 at 9:44 PM, Whit Broach <whitneybroach at gmail.com> wrote:
> (First post here and an R newbie. Glad to follow protocol.)
>
> A quick Google search of multiple lists, and a local search of downloaded
> hpc archives, doesn't show a thread focused on the symptom of snow crashing
> on Win7. I don't know where else to look, though I can tell that other
> people are finding similar symptoms.
>
> SNOW crashes or hangs the Rgui when my script tries to create processes on a
> Win7x64 box. A colleague has the same script running successfully on Ubuntu
> and WinXP. Sample line is:
>
> clust <- makeCluster( 2, type = "SOCK",
> outfile="C:/Windows/Temp/snowlog.txt" )
>
> The malfunction occurs so soon after that line of script is invoked that the
> outfile does not appear on the disk. In the Performance Monitor the Rgui
> goes to zero CPU utilization almost immediately.
>
> List searches show the most likely root causes that the community has
> identified, such as an anti-virus or firewall conflict. (No luck.) The
> community recommends including "manual = TRUE" to see how the package calls
> up subprocesses on the box, but that's where my R skill hits the wall. Even
> still, the fact that Rgui goes to zero cpu and the log file never appears
> suggests that user isolation of the problem might not even be feasible.
>
> I've also tried: removing all versions of R, deleting the R folder to take
> out the packages and then installing the latest R. The fresh installation
> "works" as far as it goes, but then the script produces the same symptom.
>
> The script has previously worked on this box, so maybe a Win7 patch blew
> away an undocumented dependency? But I would expect to see that in the
> community chatter and don't.
>
> Again, just looking for a trailhead. And glad to be a responsible list
> member.
>
> Best regards,
>
> Whit Broach
>
> Denver, CO
>
>
>        [[alternative HTML version deleted]]
>
> _______________________________________________
> R-sig-hpc mailing list
> R-sig-hpc at r-project.org
> https://stat.ethz.ch/mailman/listinfo/r-sig-hpc
>