> It's accessible from my client installs. It was working well for NFS, > I've switched to using HTTP more recently, but I've done lest testing in > that context. > > If I make an updated tree, its release number gets a letter (u for > updated). > > Extras contains additional packages I want to install. ks contains > kickstart files. dist contains the contents of the ISOs, created as per > the instructions on the CD. updates is a mirror of the relevant RHL > updates tree. > > I have an Apache virtualhost configured for the area, pointing at > /var/ftp/pub/ which is also my nfs export (ro). > > When I create an updated tree, I prefer to do it on another machine to > ensure I don't clobber anything important. Here is the gist of my maintenance mechanism: 1. copy the rpms I want to add/update the tree with to some dir, so they're all together. 2. I normally check sigs, sign them, etc, at this point 3. then I run add-rpm and tell it to test them for addition to the tree 4. if all tests out sanely then it copies the new rpms to a subdir at the same level where RedHat is. Then copies the old ones out to /tmp and symlinks the new ones into place. 5. Then I run a second script which depchecks the tree (using yum) rebuilds the yum headers, and builds the hdlist. 6. Then I rsync it to the distribution http/ftp server for campus. Does that make any sense? -sv
