How do I go about upgrading an old gentoo box?

[1clue]

Hi,

I'm not asking for specific help with my system, although I have a system I'm describing that I want to upgrade. I'd like guidance but I want to do the actual work myself.

Every so often somebody (it has been me) will post a question about a system that's been turned off for 6 months or so, and now they want to update it to current. Somebody comes back with a list of dates to update portage to, where they go to that specific date and then upgrade/recompile. Then they go to the next point. Described here: https://wiki.gentoo.org/wiki/Upgrading_ ... er_systems and here: https://wiki.gentoo.org/wiki/Upgrading_ ... ld_systems

What I'm asking is how to decide what points to update to for my rebuilds to work? How do you decide? The howto says every 4 months or so, but is there a way to tell when a significant event happened where your tree needs to be first updated to a point within a week of that event, then get just past the big jump and build again? Or are we just putting on the blinders and trusting it will all work out?

What I'd like is to be able to look at the portage timeline, see that a Big Event happened and then walk softly while I try to sort it out.

Thanks.

[Anon-E-moose]

I've upgraded a several year old system before.

For me it wasn't bad, I had daily backups of the portage dir and all tarballs that I've ever downloaded.
I started slow, jumping 3 months at a time, and then the last year 6 mo jumps.
No problems with conflicts, maybe luck, maybe 3 months is not too long for major problems.

Since I had daily backups, if I had found problems emerging then I could always delete portage (minus distfiles/packages) and load a month earlier snapshot.
Edit to add: to clarify when I said "problems emerging" I meant unsolvable conflicts before anything actually starts emerging. I always do a pretend emerge anyway.

[NeddySeagoon]

1clue,

I wrote HOWTO Update Old Gentoo

Several users have followed it. Ignore Plan A ... but i just had to try it.

[Hu]

Mostly, the big events are general knowledge on the forums because they caused notable numbers of less experienced users to turn up seeking help, or are remembered because they were painful even for experienced users. As for how to pick the right points in history, the easiest solution is to optimistically jump as far as you think might work. If you run into major failures (dependency conflicts, hard failures, etc.), pick a less optimistic point, move to that point in history, and try again.

Also, remember the power (and danger) of the big hammer approach: if you only require that the system be functional at the end, but can tolerate some impaired functionality during the process, removing non-critical packages can make your life much easier because it will reduce the complexity of the dependency tree, and may save you building updates that will be obsoleted by later updates from a tree that is closer to the present. In severe cases, this might mean that the system has no working Desktop Environment for the duration. Whether that is an acceptable trade-off is a situational decision. Some people who need to do these big updates also have a strong reason for requiring the system to be as functional as possible in the interim (such as it being the family media device, or the primary work/personal computer). Some don't (such as if the system has been powered down for 6 months already, so no one will miss it being unavailable for a few days more).

[NeddySeagoon]

1clue,

The big events also have news items
You can read them at that link. Provided your install is post GLEP42, portage will warn when you are affected by a news item.

[1clue]

1clue,

I wrote HOWTO Update Old Gentoo

Several users have followed it. Ignore Plan A ... but i just had to try it.

There are still CVS mirrors?!!!?

[1clue]

Mostly, the big events are general knowledge on the forums because they caused notable numbers of less experienced users to turn up seeking help, or are remembered because they were painful even for experienced users. As for how to pick the right points in history, the easiest solution is to optimistically jump as far as you think might work. If you run into major failures (dependency conflicts, hard failures, etc.), pick a less optimistic point, move to that point in history, and try again.

This is exactly what I'm trying to avoid.

Also, remember the power (and danger) of the big hammer approach: if you only require that the system be functional at the end, but can tolerate some impaired functionality during the process, removing non-critical packages can make your life much easier because it will reduce the complexity of the dependency tree, and may save you building updates that will be obsoleted by later updates from a tree that is closer to the present. In severe cases, this might mean that the system has no working Desktop Environment for the duration. Whether that is an acceptable trade-off is a situational decision. Some people who need to do these big updates also have a strong reason for requiring the system to be as functional as possible in the interim (such as it being the family media device, or the primary work/personal computer). Some don't (such as if the system has been powered down for 6 months already, so no one will miss it being unavailable for a few days more).

The real-world system in question is pretty minimal right now, but there are some things I wouldn't mind removing. Thanks.

[1clue]

1clue,

The big events also have news items
You can read them at that link. Provided your install is post GLEP42, portage will warn when you are affected by a news item.

Like so many of your posts, this one is truly helpful.

Thanks.

[NeddySeagoon]

1clue,

Yes. Gentoo CVS is still there.
You can get all the ebuilds that were ever in the ::gentoo repo.
How old are we talking about?

Getting old sources ... depending how far you need to go back, is the main issue.
A few readers here (me included) never purge distfiles, so I have all my distfiles online back to April 2010.
I can get back further if I fix my old box.

[krinn]

my rules for upgrading old system are simple:
1 - never cares about world : there's just no point in updating world
2 - cares about system only
3 - cares about tools that are use to upgrade and compile anything (so i care about portage, python, gcc, binutils...)

in real so i goes backward in the list, i do 3, then 2, then once 2 is all update, i finish with 1
i would go with quickpkg before critical updates attempt (the tools from the list in 3)
and my strategy is, try update portage to latest, try update to portage to a version in between if you cannot reach latest portage directly (it happen when you need a transitional portage that won't complains about ebuild API version)

i think caring about world is what stop the most the users, because they try and gets lot of errors or blockage for packages they have no use (who the hell need an update xorg server to compile?), so rather then removing them to ease the udpate, i simply don't update them until i have the tools upgrade to try update them correctly

[Rad]

I too have done this a few times before. Basically ensure that you have busybox with all utilities and a live USB with some rescue distro just in case, then just go for updating Python & Portage first. Resolve blockers mostly by uninstalls. If it actually breaks in the process, try fixing it in this way or any other that applies. After that, basically just update @system and then the rest of @world.

It can theoretically help to have Sabayon's package manager or other portage binary repositories available, but only if you're already familiar with them. Simply fixing Gentoo if things break is likely quicker otherwise.