On 08/13/2016 10:22 AM, hw wrote:
james schrieb:
On 08/12/2016 07:26 AM, hw wrote:
Michael Orlitzky schrieb:
On 08/10/2016 06:54 AM, hw wrote:
Hi,
I´m trying to upgrade portage because I´m getting a message that it
needs to be able to work with EAPI 6 packages and can only do EAPI 5.
I´m running into merge conflicts when trying to update portage, and
apparently one of the packages (dev-python/cryptography) I could try
to update first to be able to update portage requires a version of
portage that can handle EAPI 6 packages.
Try the other suggestions first -- but as a last resort -- you can
always grab a new stage3 that should contain an updated version of
portage and simply overwrite the portage files on your machine. A
quickpkg from another Gentoo machine (or the liveCD?) would also work.
I´m trying to update a production server here. If I overwrite the whole
system, who knows what might break. I can take it down for a few
hours in the evening unless I want to work over night, which is not
really an option.
There must be a way to update a Gentoo installation without breaking it.
As wonderful as it otherwise is, updating Gentoo is always a nightmare
which
makes me very seriously consider not to use it anymore. Updating
needs to
be easy and flawless and not something you always run into weird issues
with.
When I run gentoo as a critical server, I always have a second,
redundant system pretty much identical, on stanby. I upgrade the
stanby first and run it a few days, then the production system. It
makes reliability extraordinarily high. But, the again, I do a version
of the
same thing with all critical systems, or I do not work on them.
This isn´t really an option because the problem is with the updating
itself, not with something not working after the update has been
performed.
If you have an identical, second system, then you can test whatever you
want on trying to complete your upgrade. Get the second system up and
running. Disconnect the ethernet on one and upgrade the other. If you
fail, then just switch the ethernet cable. Sometimes you can just
reassign the ethernet on one, so you can ssh into it and look around,
while working on the other. (2) indentical systems allow you to work on
one and keep the other stable, until the upgrade is finished.
Granted,
as successful consultant, I have that luxury.
It´s time consuming ...
Did they recently make a new liveDVD?
A few months ago.
https://wiki.gentoo.org/wiki/Project:RelEng/LiveDVD/20160514
Cool, the old one was really getting old.
Perhaps a better solution is to make a stage-4, of your current gentoo
(production) system, verify that the stage-4 works by using it to
install a similar system, test and deploy. And then hack or fix the
production system, during the daytime, at your leisure?
Stage-4 gentoo systems have been around a long time. Documentation
varies and most have their own 'home spun' approach to stage-4
replicant systems, backups etc etc.
It might be a way to try and see if upgrading from older states step by
step until the installation is current would work. But how do I do that?
In normal times, you could just make a stage 4. It's like a complete
backup that installs. Like most other similar approaches, I always test
them out to ensure they install and populate as needed. If you google,
for gentoo : stage-4 you'll find lots of examples with slightly
different steps.
Here's an old one, translated::
http://translate.google.com/translate?prev=/language_tools&u=http%3A%2F%2Fgentoo-wiki.gentoo.ru%2Fwiki%2FStage4
There are better stage-4 guides, so just google for them::
http://www.tutorials.chymera.eu/blog/2014/05/18/mkstage4-stage4-tarballs-made-easy/
https://github.com/TheChymera/mkstage4/blob/master/README.md
http://gentoo-en.vfose.ru/wiki/Custom_Stage4
http://gentoo-en.vfose.ru/wiki/Custom_Stage4#Installing_from_Stage4
Exercise some caution here. This is something that should be done and
tested before you get in 'hot water'. Still, if you can duplicate what
you have working (and test it) then you can experiment on the second
indentical server, with out fear of corrupting the only one you have
working.
How you proceed is up to you. so be *careful*!!!
Considering how much time and effort it might take and that the update
problems aren´t going to go away, I have to wonder whether it´s better to
install Debian or Centos on another machine and to migrate the services.
Doing so would allow to make some improvements and to consolidate several
physical machines into one.
Either way, you should back up your data /etc/ /var/lib/portage/world
and any other info files you have or have modified, regardless of which
OS you end up running on. Think about all the mods and installs and
critical software you have customized on the system (iptables?).
Back all of that up too.
Either way, it sucks.
System admin requires routine attention to systems, including backups
and a total failure plan for critical systems, files, configs and other
valuable information. Being in the fox-hole taking on water and low on
ideas, hopefully will motivate you to have a more robust plan next time
around.
Good luck,
hth,
James
