Anyone have any tips on the best way to replicate the contents of a repository version (i.e. all packages for b124)? I've set up a local repo using the depot.py (hg clone'd) method with plenty of space to hold the packages and attempted to use various permutations of pkgrecv to pull down packages, but pkgrecv keeps exhausting the /tmp filespace before completing the republish operation. I've even tried 16GB of RAM in the machine, but it doesn't appear to be enough for the following to complete:
pkgrecv -s http://pkg.opensolaris.org/dev -d http://localhost -r entire at 0.5.11,5.11-0.124 I've even tried dividing the job into pieces by trying to retrieve only the shell of entire (not recursive) and then getting SUNWcsd, SUNWcs and babel_install/slim_install seperately or even some wildcards (*...@0.5.11-5.11-0.124), but that also causes problems because pkgrecv aborts republishing when it hits the first duplicate package on the repo. Can pkgrecv be told somehow to use space other than /tmp to store its temporary packages before republishing them or to "continue" on error? The goal here is to create a local repo that I can point a PXE auto-install to so I don't need to go outside the LAN to install a machine. Right now I get the auto-install to retrieve the fixed version shell of entire from my local repo (to force a particular build version), but it then proceeds to pull the rest of the data from the Internet dev repos which is SLOW. Considering I need to deploy this to about 20 machines next week, I need a way to speed this up. Can anyone offer pointers to a more efficient or effective approach to this? Ideally, I could download a tarball of the entire repo for just a particular build, extract to my local repo machine then point the depot.py to it...not sure if this is offered anywhere. -- This message posted from opensolaris.org
