Karsten McMinn wrote on Thu, Dec 27, 2007 at 11:21:54AM -0800: > obviously automation. regardless of personal administration ethics > it seems like a fair question.
If you understand the OP's question that way, you should also provide the following answer to the OP: There is no standard way for automated upgrades on OpenBSD. The standard upgrade procedure requires booting an install system, usually from floppy, CD-ROM or bsd.rd, and rebooting once more when the upgrade is done to get back to the production system. I'm not aware of any sensible approach to automation of this standard upgrade process. > Brad, you could crontab the cvs update on the local source tree, Combined with what follows, this is certainly bad advice. HEAD is a moving target. Sometimes, HEAD won't even compile if you hit right in between two related commits. So, installing self-compiled HEAD stuff via cron on a production system is asking for trouble. When you simply want to run -current, snapshots are recommended. > compile and install kernels And reboot from cron after installing the kernel? On a production system? I would call that scary. On the other hand, not rebooting after installing the new kernel is even worse. Some mini flag day might suffice to break part of your userland. That won't happen often, but on a production system, you probably do not want to break things even once or twice a year. You know, *if* cron brings your server down, it will very probably be right after the start of your long holiday. > and userland out of crontab however often you want. Upgrading userland from cron? I wouldn't call that impossible, but... Have a look at http://www.openbsd.org/faq/upgrade42.html Specifically, you need the section entitled "Upgrading without install kernel" starting with "This is NOT the recommended process. Use the install kernel method if at all possible!" There are several steps to perform. Some of them are not trivial, but they require thought. There is no guarantee these steps are always the same: Already the filename "upgrade42.html" is giving that away. Almost certainly, some things will change during the five years to come. So, scripting this is certainly possible, but it will be *much* more fragile than upgrading manually, keeping the scripts up to date will certainly be more work than doing manual upgrades twice a year, und it is definitely not a job for newbies. > likewise if you wanted a binary route (snapshots). A bit better, but still: - Do you reboot from cron? - How will you make cron read, interpret and act according to http://www.openbsd.org/faq/current.html? - How will you make cron keep /etc in sync with the system? Keeping a system up to date involves manual work, either a little easy work for manual upgrades now and then, or lots of hard and scary work for building and maintaining an automatic system. You choose according to your skill, and according to your time budget...
