[R-SIG-Mac] snow cluster - unable to open connection

[Luke Tierney]

I don't see this on my Mac OS X setup (10.3.something).  I'm not at
the console so can't easily check my firewall settings.

The symptom suggests that a previous attempt to open a server socket
on port 600011 has not been cleanly completed.  You might check with
netstat whether there is any information on this. As long as the
processes are no longer around this usually clears itself within a few
minutes.

I would follow Sean's suggestions on isolating this; also try your
initial machine both as localhost and by its IP address or name to try
to get it through the network interface.

Best,

luke

On Mon, 4 Sep 2006, Sean Davis wrote:

> Martin Renner wrote:
>> I'm trying to build a sock cluster using snow but have difficulties
>> to include more than one machine. In the terminal, password-less ssh
>> is working on all machines and each machine can talk to itself (cl <-
>> makeSOCKcluster ("localhost") works). Trying to cluster more than one
>> machine, I get:
>>
>> Error in socketConnection (port = port, server = TRUE, blocking =
>> TRUE,  :
>> 	unable to open connection
>> In addition: warning message:
>> port 600011 cannot be opened
>>
>> I've tried to disable the firewalls as well as including "port = 22":
>> cl <- makeSOCKcluster (c("localhost", "localhost", "machine2.local",
>> "machine2.local"), port = 22)
>> but this fails just as well.
>>
>> My set-up is as follows:
>> R version: 2.3.1
>> Mac OS versions: 10.3.9 (on dual G4 and G5 iMac), 10.4.7 (MacBook Pro)
>>
>> Has anybody else encountered this problem? I've searched the usual
>> sites with no luck. Any pointers would be much appreciated.
>>
> I haven't tried to make a socket-based cluster, but I have successfully
> made a lam-mpi-based cluster.  I ended up having to TURN OFF the
> firewall, as at least for MPI, the ports are assigned randomly, so
> knowing which is going to be assigned is not possible, I don't think,
> and I think that each process gets a separate port, if I remember
> correctly (but I could definitely be wrong).  In any case, lam-mpi
> installed "out-of-the-box" on my Mac Pro as well as on a G5 and an
> Xserver with XTools installed, so you could give that a try, as well.
>
> As for the more immediate problem, does running with just one machine in
> the cluster work?  If so, does using multiple "machines" on localhost
> work?  How about turning off the firewalls to the two machines--does
> that make a difference (don't do this if you don't think it is safe,
> but....)?
>
> Sean
>
> _______________________________________________
> R-SIG-Mac mailing list
> R-SIG-Mac at stat.math.ethz.ch
> https://stat.ethz.ch/mailman/listinfo/r-sig-mac
>

-- 
Luke Tierney
Chair, Statistics and Actuarial Science
Ralph E. Wareham Professor of Mathematical Sciences
University of Iowa                  Phone:             319-335-3386
Department of Statistics and        Fax:               319-335-3017
    Actuarial Science
241 Schaeffer Hall                  email:      luke at stat.uiowa.edu
Iowa City, IA 52242                 WWW:  http://www.stat.uiowa.edu