Erik LaFever wrote: > Shawn Walker wrote: >> Erik LaFever wrote: >>> DC team, >>> >>> While reviewing the results of the python26 DC testing I have 11 >>> new test failures that are all related to logging and I still see the >>> problem with pkg_image size increased from < 5mb to > 80mb. >> >> As I previously noted, the catalog data for a publisher is *much* >> larger now than it used to be. It is expected that you would see >> image size go from < 5MB to > 80MB when you have the 'opensolaris.org' >> publisher configured using 'http://pkg.opensolaris.org/dev', as an >> example. >> >> Please check the size of '/var/pkg' to confirm, and remember that this >> information compresses extremely well. >> >> Cheers, > If I create a pkg_image with the --no-refresh option, I get a size under > 5mb. If the --no-refresh option is not used, I would see a size over 80mb. > DC (up to build 127) created a pkg_image with a small footprint. Now > the footprint is larger (yes you mentioned that the catalog is larger) > > I don't know if anything changed in DC or if they were using the > --no-refresh option previously for pkg_image creation. > The new logging that they are doing shows the full "pkg" commands and I > do not see a --no-refresh option being sent. > > If I manually create a pkg_image and use the -no-refresh option, the > size is small. DC has historically been working to conserve space. I > have a couple of tests that use small partitions. These tests fail now > during the pkg_image creation when they used to be failing during pkg > installation.
I'm not part of the AI/Caiman/Install team, to be clear. And I saw your bit about --no-refresh, but was somewhat confused by it. An image created with --no-refresh isn't very useful. In particular, once you install any packages into it, the catalog will have to be retrieved and you're back where you started. Is there a particular reason why a DC image with no catalog data is useful? Cheers, -- Shawn Walker
