Extracting data from a raster brick using multiple threads

Question

I'm extracting data from a raster file. I need to extract only a random sample of 1000 cells, like so:

pr <- brick(filename[i], varname=var[i], na.rm=T)
cells <- sampleRandom(pr[[1]], 1000, cells=T)
prvals <- pr[cells[,1]] #Equivalent of extract(pr, cells[,1])

I'd like to use more cores to speed up the process, like so:

beginCluster(4)
pr <- brick(filename[i], varname=var[i], na.rm=T)
cells <- sampleRandom(pr[[1]], cnsample, cells=T)
prvals <- clusterR(pr, extract, args=list(cells[,1]))
endCluster()

However, the clusterR() line is throwing all kind of errors at me, such as:

prvals <- clusterR(pr, extract, args=list(cells[,1])) Error in checkForRemoteErrors(lapply(cl, recvResult)) : 4 nodes produced errors; first error: argument is of length zero

prvals <- clusterR(pr, extract, args=list(cells[,1])) Error in checkForRemoteErrors(lapply(cl, recvResult)) : 2 nodes produced errors; first error: argument is of length zero

prvals <- clusterR(pr, extract, args=list(cells[,1])) Error in checkForRemoteErrors(lapply(cl, recvResult)) : 3 nodes produced errors; first error: argument is of length zero

prvals <- clusterR(pr, extract, args=list(cells[,1])) Error in clusterR(pr, extract, args = list(cells[, 1])) : cluster error In addition: There were 11 warnings (use warnings() to see them)

prvals <- clusterR(pr, extract, args=list(cells[,1])) Error in checkForRemoteErrors(lapply(cl, recvResult)) : 3 nodes produced errors; first error: argument is of length zero In addition: Warning message: closing unused connection 9 (/tmp/R_raster_tmp/afantini/raster_tmp_2014-11-28_131043_31528.gri)

What's wrong?

EDIT: I tried the examples in the raster documentation and they work on another machine, on this they don't. So there is a deeper problem with this machine. Package snow is correctly installed. R is at latest version.