How to configure MLS/RGUI to use local repo? RRS feed

  • Question

  • My organisation has a new implementation of MLS installed on-premise on Windows servers. Our architecture consists of 5 servers. One server is configured as the header node, and four servers are configured as compute nodes. Our users create a remote session to the header node using RGUI installed on their local desktop.  All of the servers are located behind our DMZ and do not have access to the www.

    We are successfully using miniCRAN to download the R packages that we need (plus all of their dependencies).  The packages are first downloaded on an IT admin's local desktop and saved to that person's local R package repo.

    The IT admin then copies the deployable package to a shared NAS store that he can access when logging on to the four compute nodes.  The IT admin logs onto each compute node in-turn to use miniCRAN to deploy the package to the default local repo.  In our case "C:\Program Files\Microsoft\ML Server\R_SERVER\library".

    To simplify the package deployment process we want to create a centralised package repo on a NAS share that all four compute nodes have access to.  (This will save having to deploy to all 4 compute nodes in turn)

    How do I configure MLS or RGUI so that it refers to our new centralised package repo and not the default library created on the server C:\ drive during the MLS s/w installation?

    (I.e. what files do I have to update and what commands/parameters do I use/assign?)

    Friday, June 14, 2019 2:38 PM

All replies

  • You will need to map the NAS share to a local drive letter on each of 4 compute nodes. Then you can use the R function .libPaths() to configure the default R library directory and set that in the file that gets included in the default MLS install(C:\Program Files\Microsoft\ML Server\R_SERVER\etc\

    For an example of how to do this you can take a look at this post:

    Hope this helps.

    Friday, June 14, 2019 8:56 PM