> Consider a binary server. Build on one machine, install on many. > > emerge -gK world > > will use the binary server to intelligently fetch updates > and use HTTP/FTP for the process. It will merge extactly > what is posted on the server based on the metadata in the > binaries. > > --NJ >
I use this setup all the time. I maintain 14 Gentoo boxes, seven of which are off site using this method. I also have setup a portage mirror on the binhost using gentoo-rsync-mirror that updates three times a day that I sync the other 13 boxes to. I run emerge -uDf on the binhost after the rsync is completed so it will fetch packages ahead of time on my binhost. Before updating any other machines, I ssh into the binhost and run an emerge -uDb world and build binary packages for updates. For fetching packages from the remote machines, I use apache and foward port 8080 to an apache instance on my binhost for package retrieval. Really simple system. I'm sure there may be better systems, but this is fast and effective. It conserves bandwidth for the gentoo project, and gives me fast updates for all the machines I have to maintain. It rarely takes more than a half hour to do an update on any given machine. I also use distcc on the seven local boxes at my office. Compile times are short. I have occasionally had problems with this setup, but that has always been attributed to changes in slotted package names. Or packages being moved to slotted. I think these problems are worked out for the most part, so this shouldn't be an issue. HTH Wendall -- [email protected] mailing list
