The higher-ups at our institution (MD Anderson Cancer Center) have made the decision to disable Internet access on our computing cluster, and only allow incoming access to specific IP addresses.

In order to install Bioconductor packages, we need a list of stable IP addresses for Bioconductor that we can whitelist. However, the IP address for bioconductor.org does not appear to be stable (currently 13.33.112.5, pointing to an Amazon server).

This question was asked before, but the provided answer is no longer applicable, as the recommended stable IP address is no longer hosting a Bioconductor mirror (140.107.3.20; Fred Hutch).

I suppose one solution is to whitelist all of AWS's IP ranges... but this might not be acceptable to the higher-ups.

Does anyone have other solutions to this firewall problem?

EDIT:

As Martin indicates below, master.bioconductor.org has a stable IP of 52.71.54.154 and is serving packages. Perhaps options(BioC_mirror="52.71.54.154"); biocLite("DNAcopy") will work once 52.71.54.154 is whitelisted on our network.

Will it be acceptable to Bioconductor for us to bypass load balancing and use 52.71.54.154 as the sole IP address for downloading packages on our Internet-disabled computing cluster?

master.bioconductor.org is in the Amazon cloud, but 52.71.54.154 is a static (elastic) IP address; it is stable.

An alternative is to clone, using rsync, the Bioconductor (and/or CRAN) repository and arrange for your cluster to point to the local repository (via R.home("etc/Rprofile.site"), for instance, with options(Bioc_mirror = ...), see ?setRepositories).

Many clusters have a 'head' node that has access to outside IP addresses; the requirement then is to install packages on the head node.