[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Administration (+apt-get dist-upgrade) of 100s of machines



Daniel Palmer daniel-at-cardboardbox.org.uk |volatile-lists| wrote:
Jamming dist-upgrade into a cron job will cause problems when a package doesn't upgrade cleanly.. for example mysql is getting upgraded, the server will stop and not come back up. Even worse if a kernel upgrade doesn't create the initrd correctly.. basically too much is going on and someone should be checking on it.

But that could be a problem also with just plain "apt-get upgrade", couldn't it? In both these cases apt-get will fail and $? will be set to non-zero, at least in my experience. That would work fine for me, since it allows me to detect an error situation.

I don't think apt is meant to be automated. Updating package lists and pre-caching the packages you'll need later with cron-apt speeds things up though. The best option would be to use one of the applications other people have mentioned to *manage* mass updates. The ideal solution is to sit down once a week, manage an update and have the result propagate to your machines. Looking at the screenshots, fai seems to do exactly this. RHEL has a web-based system that does exactly the same but at a price. :P

By looking at the source code for "fai softupdate", I can see that in the end, it ends up calling

"""
export aptopt=
$ROOTCMD apt-get $aptopt -f -y dist-upgrade </dev/null
"""
So I guess "fai softupdate" will have exactly the same shortcomings as putting "apt-get -f -y dist-upgrade" in a script. Then I might as well do that.

Ok, guys, I thank you for your suggestions. I just thought I had misunderstood something and that "there had to be a better way". It seems there isn't.

Its apt-get upgrade and live with the kept-back packages or dist-upgrade and deal with the potential "package breakage" - whatever that means...

Peter

--
Peter Valdemar Mørch
http://www.morch.com



Reply to: