Reputation: 1
I want to run a code on a cluster of several PCs using as current directory a common logical unit, that I called Q:\
. The correct path to this folder is, say,
\\myserver\myfolder
so I'm starting the parallel script with
spmd
cd('\\myserver\myfolder')
end
and, to be sure, I even put the folder in "AdditionalPaths" in the Properties of the Cluster Profile.
To check whether the functions and the files are reachable I use
spmd
check('function1')
end
where "function1" is a function m-file in "myfolder". Unfortunately spmd
can't find the function on every PC, but just on a few of them, with no apparent reason of choice. I already checked in the issued computers if the "SYSTEM" user account (used by matlabpool) has permission to "\myserver\myfolder", but it always has.
Anybody encountered a similar situation?
P.S: I'm using on every computer in the cluster R2012a on Windows 7 64bit.
Upvotes: 0
Views: 67
Reputation: 366
The 'additionalPaths' setting in the cluster profile applies to cluster jobs created and associated with that profile. A job needs to be created using 'parcluster(cluster_profile)' per this example, where cluster_profile is the name of the cluster profile that you have assigned the additionalPaths
variable.
Also for robustness you can try also adding the AdditionalFiles
option with the .m files needed and the path command can be added to the startup files for the worker (e.g. jobStartup.m
, taskStartup.m
) to ensure that every worker can see the path.
Upvotes: 1