Can I directly upgrade (update?) from my current 6.5 box to version 7?
Or, must I wipe my drives and install from scratch?
I wouldn’t. redhat does have an upgrade script for headless server-only type configurations, but there’s just so much room for
‘wrong’, with newer versions of all the system components like python, apache, libc and other libraries, PostgreSQL database server, etc etc etc.
I would save /home, a copy of /etc/passwd,shadow and any other specific configuration files, dump any databases, apache configurations, etc etc, then do a clean install and piece that stuff back together.
As John says, it’s risky. And you’re going to want backups anyways if things go south.
There is a tool in the works, but last I heard it needs testing. http://wiki.CentOS.org/TipsAndTricks/CentOSUpgradeTool
(I’m a member of the fresh install camp!)
Huh. I thought preupgrade was part of RHEL 7, since it’s in fedora 19 (and pretty much works, most of the time…).
Another idea: back up your o/s: mv /boot/* to /boot/old, stuff off / to
/old, then just install. Easy downgrade path….
I have try to upgrade CentOS 6.5 to 7, appears this error:
]# yum install redhat-upgrade-tool preupgrade-assistant-contents Loaded plugins: fastestmirror, presto Loading mirror speeds from cached hostfile
* base: CentOS.xpg.com.br
* extras: CentOS.xpg.com.br
* updates: CentOS.xpg.com.br Setting up Install Process Resolving Dependencies
–> Running transaction check
—> Package preupgrade-assistant-contents.noarch 0:0.5.14-1.el6.CentOS
will be installed
–> Processing Dependency: preupgrade-assistant(x86-64) >= 1.0.2-14 for package: preupgrade-assistant-contents-0.5.14-1.el6.CentOS.noarch
—> Package redhat-upgrade-tool.noarch 1:0.7.22-3.el6.CentOS will be installed
–> Processing Dependency: preupgrade-assistant >= 1.0.2-4 for package:
–> Finished Dependency Resolution Error: Package: 1:redhat-upgrade-tool-0.7.22-3.el6.CentOS.noarch (upg)
Requires: preupgrade-assistant >= 1.0.2-4
Error: Package: preupgrade-assistant-contents-0.5.14-1.el6.CentOS.noarch
Requires: preupgrade-assistant(x86-64) >= 1.0.2-14
Em 09-09-2014 14:30, email@example.com escreveu:
Em 09-09-2014 15:08, John R Pierce escreveu:
I’m using exactly this tutorial.
I didn’t see the special repo mentioned there on your yum output. re-read section 1.1.
Em 09-09-2014 15:38, John R Pierce escreveu:
I’m using this repo, see the output of yum:
Error: Package: 1:redhat-upgrade-tool-0.7.22-3.el6.CentOS.noarch
Do you have any of these lines in either the /etc/yum.conf OR
Or are you using yum priorities?
preupgrade-assistant-1.0.2-36.0.1.el6.CentOS.x86_64.rpm < == this exists in the repo
I just checked the instructions and the repo and it works:
Package Arch Version Repository Size
0.5.14-1.el6.CentOS upg 917 k redhat-upgrade-tool noarch
1:0.7.22-3.el6.CentOS upg 85 k Installing for dependencies:
1.1.26-2.el6_3.1 base 561 k openscap x86_64
1.0.8-1.0.1.el6.CentOS.1 updates 2.9 M
7.8-6.el6 base 318 k preupgrade-assistant x86_64
Johnny Hughes wrote:
Or is the repo not enabled (enabled=0)?
Em 09-09-2014 18:16, firstname.lastname@example.org escreveu:
No exist this lines in yum.conf enabled=1
Thanks a lot guys.
I just ran the pre-upgrade, and it said that I couldn’t do an in place upgrade for the following reason:
“/usr directory is on separate partition. In-place Upgrade is NOT possible.”
Well that was news to me because I never separate /usr, so what is going on? It apparently turns out that /usr/local is on a separate partition and the software isn’t smart enough to realize that there’s a big difference between /usr and /usr/local, so I’m guessing that that is the reason.
I sure do hate software that makes lame or at worst incorrect assumptions, almost as bad as when you change a config file and comment out the default line and add you duplicate the line with changes, and the software is too dumb to realize what you did.
You should be able to just unmount /usr/local and take it out of fstab for the duration of the upgrade then put it back and fix the contents later. But, that sort of thing makes a bare-metal reinstall sound even more sensible since it shows how many options there are and how many they may not have considered.
Sure, I can umount /usr/local and comment it out of fstab, my comment was more about the fact that the software confused my configuration more that than anything else. But then again it’s probably best that they take the most conservative route and make folks aware of any and all possible ways that it might fail.
The report also listed a driver for an Ethernet card that was not supported in CentOS 7 which does make the software useful for identifying things like that.
All in all I think that it’s worthwhile to run the software to see what it finds so that you can be aware of the types of issues that you might run into.
I agree that doing an upgrade using this method would be a bit of a last resort, I might try it just to see how it goes, just for the heck of it, but ultimately a re-install is best as it get’s rid of all the chaff as well.