Feature request: Support for Parallel computing toolbox general scheduler with mounted storage

1 次查看(过去 30 天)
I have a few convinient shortcuts on the Matlab client computer, for which I want Matlab to take advantage of:
  • Mounted storage via sshfs: I would like the Matlab client to access it as a shared generic scheduler. However the location on the client and server do not match, and RemoteJobStorageLocation is not read for remote or shared generic schedulers, so I get errors when trying to submit a job with a remote generic scheduler. Allowing for the following code in independentSubmitFcn would allow for such a setup, however this is not possible because of the failsafes introduced on Matlab's side.
%% independentSubmitFcn
% The local job directory
localJobDirectory = cluster.getJobFolder(job);
% Find out how we should refer to the job storage location on the cluster.
% remoteJobDirectory = cluster.getJobFolderOnCluster(job);
remoteJobDirectory = remoteConnection.getRemoteJobLocation(job.ID, cluster.OperatingSystem);
  • Aliased scheduler controller (e.g. in Slurm: alias sinfo=ssh login_node sinfo): I am not sure if there are any problems that could pop up from using this approach, but if possible along with the changes above, it would be possible to directly setup the cluster profile offered by Matlab.
  • ssh-agent managed connections: I would like to take advantage of the paswordless connection offerred by the ssh-agent and not have to input a user password or expose the location of an identity file in the matlab configuration. Similarly if the ssh connection can use pre-configured setup in .ssh/config that would help as well.
Maybe some of these issues already have documented sollutions, in which case I would love to know about them, but I would still like to know the development team's feedback on introducing these options in the base code. If there are any issues with such setup I would like to know before diving deeper into implementing this solution.

回答(0 个)

类别

Help CenterFile Exchange 中查找有关 Third-Party Cluster Configuration 的更多信息

产品


版本

R2020b

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by