On Sat, 2008-01-19 at 16:36 -0500, seth vidal wrote:
> On Sat, 2008-01-19 at 16:18 -0500, Konstantin Ryabitsev wrote:
> > On Jan 19, 2008 4:01 PM, seth vidal <[EMAIL PROTECTED]> wrote:
> > > You're not wrong, I was just describing what someone had told me they
> > > had done with a yum plugin and it was highly similar to the apt+zfs
> > > article. sorry for the confusion.
> > >
> > > LVM is not unique to the linux implementation. Several lvm mechanisms
> > > (on things like veritas for example) allow snap file system copies.
> > 
> > Needless to say, if there was a way to do safe FS-level RPM rollbacks,
> > lots and lots of people would be immensely pleased with that. :)
> > 
> > Of course, that is still rife with plenty of ways it can go horribly
> > wrong. E.g. what happens when a %post restarts a process, and then we
> > roll back the fs? Tricky stuff.
> 
> Not at all. You make it a complete system-state rollback. Take a
> snapshot of system memory too. :) I mean software-suspend to disk is all
> about a snapshot of memory at that time, right?

 Well the article seemed to be using reboots as a way to "rollback",
suspend would just be a "fast reboot".

 However personally I don't think it's worth much, there's just too much
collateral damage ... for instance it's not uncommon to have / and /var
on the same partition, so you'd be rolling all your log files back and
mail/news/HTTP updates.
 And if you are going to have that much pain, do it via.
virtualization ... at least then you can run both the "old" and "new"
machines at the same time.

 Also generally everyone seems to think rollbacks are a really hard
technical problem, because they want to solve for the kernel %post or a
glibc update going wrong etc. on the first attempt.
 But I think they are a simple technical problem, as I'm pretty sure 90%
+ of the requests for rollbacks go away when/if we can easily say update
"vim" or something else simple-ish[1] ... and then have some UI where
they say "vim" is buggy, revert. And it just re-installs the old
version.
 For instance Fedora has a huge amount of updates, but a lot of the
bigger ones (the harder ones to rollback) don't change as often and get
a lot more testing.


[1] vim happens to have no scripts, which is actually less common than
you might think due to things like gconftool-2 and
gtk-update-icon-cache ... but I think we could work around the simple
ones too.

-- 
James Antill <[EMAIL PROTECTED]>
Red Hat

Attachment: signature.asc
Description: This is a digitally signed message part

_______________________________________________
Yum-devel mailing list
[email protected]
https://lists.dulug.duke.edu/mailman/listinfo/yum-devel

Reply via email to