What is CentOS?: CentOS is essentially a rebuild of the RHEL source code. I've been using this OS for a couple weeks now with no issues running cpanel.
More info on CentOS can be found on the web site:
http://www.centos.org

Here are the steps i took to update a Redhat 9 box running on a minimal install ( no X, etc ) as suitable for a server to CentOS 3.1 ( which is a fork of RedHat Enterprise Linux v3 )..

I took a fresh install of redhat 9, slapped it on a test box then updated it with up2date. I then installed yum for redhat 9 which can be found here:
Yum Download (http://linux.duke.edu/projects/yum/download.ptml)

Now, you might want to rebuild the rpm database, just incase, thats up to you:
# rpm -vv --rebuilddb

I then modified /etc/yum.conf to look like this:


[main]
exclude=httpd* mysql* php* perl*
cachedir=/var/cache/yum
debuglevel=2
logfile=/var/log/yum.log
pkgpolicy=newest
distroverpkg=redhat-release
installonlypkgs=kernel kernel-smp kernel-hugemem kernel-enterprise kernel-debug kernel-unsupported kernel-smp-unsupported kernel-hugemem-unsupported
tolerant=1
exactarch=1

[base]
name=CentOS-$releasever - Base
baseurl=http://mirror.pacific.net.au/linux/cAos/centos-3/3.1/os/i386/
gpgcheck=1

#released updates
[update]
name=CentOS-$releasever - Updates
baseurl=http://mirror.pacific.net.au/linux/cAos/centos-3/3.1/updates/i386/
gpgcheck=1

#packages used/produced in the build but not released
[addons]
name=CentOS-$releasever - Addons
baseurl=http://mirror.pacific.net.au/linux/cAos/centos-3/3.1/addons/i386/
gpgcheck=1

#additional packages that may be useful
[extras]
name=CentOS-$releasever - Extras
baseurl=http://mirror.pacific.net.au/linux/cAos/centos-3/3.1/extras/i386/
gpgcheck=1


Change "http://mirror.pacific.net.au" to a mirror near you, a list can be found here:
CentOS Mirrors (http://www.centos.org/index.php?option=com_mirror2&Itemid=84)

After that you need to import the GPG key for the centos RPMs. Issue the following command:
# rpm --import http://mirror.centos.org/centos/3.1/i386/RPM-GPG-KEY-CentOS-3

Nows the time you want to check you have made appropriate backups, etc ( i hope you all test this out before rolling production servers! ).

Next step is to run the following command:
# yum upgrade

This will download the rpm headers and determine what needs to be done, play carefull attention of course here and in all parts if this process for errors. Yum will then ask you if you are sure you want to go ahead, if no errors are present, press "y" then "enter".

This process will take a while, depending on your connection and the gear your installing it on of course. Go brew a fresh pot of coffee, order a pizza, grab a movie or GET BACK TO WORK YOU!!!!

After thats done, check for errors, make sure the new kernel is installed by issueing:
# rpm -qa|grep kernel

You should be presented with a kernel versioned like this: 2.4.21-15.EL.c0 as well as your other kernels previously installed. Check your boot loader to make sure the new kernel is default.

Finally run:
# yum update

Just to ensure you have everything you need.

Now your set to reboot ( hopefully ):
shutdown -rf now

This worked like a charm for me, hope you all can make use of this somehow :)

its little hard to do thast i think

Its fairly straight forward actually, what are you stuck on?

yes looks very stright forword

[root@opium root]# uname -r
2.4.21-15.EL.c0
[root@opium root]# cat /etc/*release*
CentOS release 3.1 (final)
[root@opium root]#


Here's a direct admin box upgraded from rh9

thnx TLG for testing that out.. ur a legend!! :)

And just incase, cuz i've had a couple people have this issue now, if the EL kernel doesn't install ( rpm -qa | grep kernel <- to check + cat grub.conf ; else lilo.conf ), make sure you download the rpm and rpm -ivh kernelfilename.rpm and modify your boot loader ( be it lilo or grub ) before you reboot! The rpm can be found on any of the CentOS mirrors.

The upgrade seems very stable. On a side note, it does not break anything with directadmin. I did not have to fix anything. I did however have to add a few http related thing to the skiplist and webalizer.

Originally posted by thelinuxguy
The upgrade seems very stable. On a side note, it does not break anything with directadmin. I did not have to fix anything. I did however have to add a few http related thing to the skiplist and webalizer.

Man I'm glad I came across this post on google, I wanted the Redhat 3. es and this release does it just as good!

now to play around with Tomcat and Java

has anyone tried this on cpanel boxes?

So far i've done 1 personal production redhat 9 cpanel box no problems. That box also had ircd's and eggdrops on it. No reported problems.

I've been in contact with others that have successfully migrated their cpanel rh9 to centos via this howto as well with no issues.

A good yum.conf for the USA

[main]
exclude=httpd* mysql* php* perl*
cachedir=/var/cache/yum
debuglevel=2
logfile=/var/log/yum.log
pkgpolicy=newest
distroverpkg=redhat-release
installonlypkgs=kernel kernel-smp kernel-hugemem kernel-enterprise kernel-debug kernel-unsupported kernel-smp-unsupported kernel-hugemem-unsupported
tolerant=1
exactarch=1

[base]
name=CentOS-$releasever - Base
baseurl=http://caos.nplus1.net/centos-3/3.1/os/i386/
gpgcheck=1

#released updates
[update]
name=CentOS-$releasever - Updates
baseurl=http://caos.nplus1.net/centos-3/3.1/updates/i386/
gpgcheck=1

#packages used/produced in the build but not released
[addons]
name=CentOS-$releasever - Addons
baseurl=http://caos.nplus1.net/centos-3/3.1/addons/i386/
gpgcheck=1

#additional packages that may be useful
[extras]
name=CentOS-$releasever - Extras
baseurl=http://caos.nplus1.net/centos-3/3.1/extras/i386/
gpgcheck=1

Taken from myCPAdmin.com (http://mycpadmin.com/index.php?showtopic=7)

Would this work for Fedora to Centos too?

Simple Answer: No! This how-to will not work with fedora as it is much newer than centos 3.1.

As fedora has newer technologies ( newer gcc, kernel, etc ) this would be a very trickey transition. In essence, you will be *downgrading* the box. I have not had the chance to test this as we only use 1 fedora box that we are migrating clients to a brand new box shortly, thus I have not personally had the need to do so.

Complicated answer: I know it has been done, but just what needs to be done to enable a smooth transition I am unsure of. You may want to contact thelinuxguy on these boards or through his web site: http://www.rack911.com as I am sure he's told me he has done it with success.

Hm that's too bad. The reason that I wanted to try it is that CentOS is much less of a beta product than Fedora is. The problem here would be that I'd have to upgrade the OS on about 50 machines every time they release a new version..

if i had another test box with me i could make a tutorial probably.

Haze, did you try with:
installonlypkgs=kernel*
Instead of:
installonlypkgs=kernel kernel-smp kernel-hugemem kernel-enterprise kernel-debug kernel-unsupported kernel-smp-unsupported kernel-hugemem-unsupported
I tried the wildcard which also works for my test box just FYI and maybe others can try :p

Choon,

That'd probably work just fine, infact, i see no reason why it wouldn't. I just basically used a default config that was either provided by yum-conf rpm or was modified by one of cpanels checkyum or safeyum scripts. At the time, since its a rather big transition, i just thought it'd be better to be safe than sorry.

Hi,

I was following the instructions and the following came up:

.....Unable to satisfy dependencies
Package mod_python needs httpd >= 2.0.40, this is not available.
Package mod_python needs httpd-mmn = 20020903, this is not available.
Package mod_ssl needs httpd = 2.0.46-32.ent.3.centos.1, this is not available.
Package mod_ssl needs httpd-mmn = 20020903, this is not available.

I am guessing i will need to update apache to 2.0.46 and higher for this CentOS configuration to work.

Is that correct?

Hrm.. is there any perticular reason you have those modules installed in the first place ?

Also, are you using apache from source or from rpm ? If you build apache from source, generally, you build modules from source anyways is why I ask.

This server is a stock cpanel server. So those modules are what came with the server when i took over it.

Well.. if you don't need them, why have them installed? Its wise to ensure you don't have anything installed you don't need. Especially with a new server and even moreso if you've been handed a server from someone else. A lot of linux distro's install a lot of programs that you will find useless especially with cpanel installed. So, seeing as your doing such a large move, you'd be best of to sort through what rpms you have installed and what you don't need. That way your upgrade will be a hell of a lot faster and less prone to problems such as you have already experienced.

just rpm -e all the dependencys and continue installing.

Thanks thelinuxguy,

Thanks all for this great howto and guidance. It was so easy to do. It worked without a hitch. Just need to test it.

Thanks again.

Just make sure not to rpm -e perl (it came up as a dependency once for me), so I removed it, but that messed up the whole yum script ofcourse..

Instead of rpm -e mod_python you could also add it to the exclude list..

Adding it to the exclude list is not a good idea, it wont be updated and you could have a vulnerable version installed.

For WHM/cPanel, do the following to check:
rpm -qa|grep httpd
rpm -qa|grep mod_
rpm -qa|grep php
Basically any output from the above commands... those packages can be removed not skipped... just my thoughts ;)

Oh sorry, I put that a bit in a wrong way.

I was talking about my situation: I currently have fedora core 2 installed and all software that I have installed is newer (or the same) as what CentOS 3.1 has, as it is older.

So there wouldn't be a vulerable version installed, in my case..

Originally posted by Haze
Choon,

That'd probably work just fine, infact, i see no reason why it wouldn't. I just basically used a default config that was either provided by yum-conf rpm or was modified by one of cpanels checkyum or safeyum scripts. At the time, since its a rather big transition, i just thought it'd be better to be safe than sorry.
I am going to hide myself for giving false information... well to certain extend.
The wildcard didn't work as it will upgrade kernels instead of installing kernels :( I just noticed it by doing my second test using my test system which rpm -qa|grep kernel output cannot see any old kernels :p

I'm just happy i wasn't the lab rat on that one choon :) cheers.

On second thought, the following should work:

installonlypkgs=kernel* kernel-*

....Unable to satisfy dependencies
Package rh-postgresql-server needs libcom_err.so.3, this is not available.
Package openssh-server needs libcom_err.so.3, this is not available.
Package samba-common needs libcom_err.so.3, this is not available.
Package imap needs libcom_err.so.3, this is not available.
Package openssh-clients needs libcom_err.so.3, this is not available.
Package samba needs libcom_err.so.3, this is not available.
Package openssh needs libcom_err.so.3, this is not available.
Package cyrus-sasl needs libcom_err.so.3, this is not available.

I tried:

root@box [/usr/src/apf-0.9.4-5]# rpm -e rh-postgresql-server
error: package rh-postgresql-server is not installed
root@box [/usr/src/apf-0.9.4-5]# rpm -e cyrus-sasl
error: Failed dependencies:
libsasl.so.7 is needed by (installed) mutt-1.4.1-5
libsasl.so.7 is needed by (installed) squid-2.5.STABLE3-0
libsasl2.so.2 is needed by (installed) openldap-2.1.22-8
cyrus-sasl is needed by (installed) openldap-2.1.22-8
cyrus-sasl = 2.1.15-6 is needed by (installed) cyrus-sasl-md5-2.1.15-6
cyrus-sasl = 2.1.15-6 is needed by (installed) cyrus-sasl-plain-2.1.15-6

What shoudl i do if rpm -e doesnt help?

First i need more information. What OS are you moving from. What steps have you taken so far to get where you are. Have you already run yum upgrade/update ?

It seems you've removed the rpm for krb5-libs (might want to re-install this one), any perticular reason for that ?

I ran # yum upgrade on a fedora box.

It has updated many modules, but suddenlystopped here after 2 minutes..

STOP! You can NOT use this howto to upgrade to centos from fedora. Fedora is NEWER technology than centos is. CentOS is an enterprise OS with longer support than most linux distobutions. What im getting at here is that, you are essentially downgrading your OS.

That being said, there is a way for it to be done, but i have not had a chance to try it myself, so i can not provide you with assistance with this.

Do a search on google, there are how-to's available on the internet for you to use as a starting point. I would suggest however that you start from scratch using CentOS. If centos is not available to you for whatever reason, perhaps ask for Redhat 9 and move up yourself from there.

Ouch, can I undo everything now,since it did install some modules..:(

You can try reverting your yum config back to what it was previously ( or as it should be configured for whichever fc version your on ). Check to make sure that /etc/*release* is set to the appropriate version as well. then run yum upgrade. This may do it for you. No guarantee's.

Thanks.

I'll check out rack911, I hope he can do it :(

Originally posted by Haze
Complicated answer: I know it has been done, but just what needs to be done to enable a smooth transition I am unsure of. You may want to contact thelinuxguy on these boards or through his web site: http://www.rack911.com as I am sure he's told me he has done it with success.

I should have read the whole thread before starting :o

Is CentOS any better than Redhat 9 performance wise ?? I mean the loads and speed..

Originally posted by Haze
You can try reverting your yum config back to what it was previously ( or as it should be configured for whichever fc version your on ). Check to make sure that /etc/*release* is set to the appropriate version as well. then run yum upgrade. This may do it for you. No guarantee's.

According to Steve, it didnt install the centos rpms, due to the dependencies error..

After checking it, I found out that he is correct.

What I currently dont understand is how is it possible to solve the dependencies errors.

This is left:

....Unable to satisfy dependencies
Package openssh-server needs libcom_err.so.3, this is not available.
Package openssh-clients needs libcom_err.so.3, this is not available.
Package openssh needs libcom_err.so.3, this is not available.
Package cyrus-sasl needs libcom_err.so.3, this is not available.

When I try to remove the cyrus-sasl module:

@@@@@@@@@@@@@@@@@@@@@@@@@@@@@
root@box [~]# rpm -qa|grep cyrus-sasl
cyrus-sasl-2.1.15-6
cyrus-sasl-md5-2.1.15-6
root@box [~]# rpm -e cyrus-sasl-2.1.15-6
error: Failed dependencies:
libsasl.so.7 is needed by (installed) squid-2.5.STABLE3-0
libsasl.so.7 is needed by (installed) mutt-1.4.1-5
libsasl2.so.2 is needed by (installed) openldap-2.1.22-8
cyrus-sasl is needed by (installed) openldap-2.1.22-8
cyrus-sasl = 2.1.15-6 is needed by (installed) cyrus-sasl-md5-2.1.15-6
root@box [~]# rpm -e cyrus-sasl-md5-2.1.15-6
error: Failed dependencies:
cyrus-sasl-md5 is needed by (installed) openldap-2.1.22-8
root@box [~]# rpm -qa|grep squid
squid-2.5.STABLE3-0
root@box [~]# rpm -e squid-2.5.STABLE3-0
root@box [~]# rpm -qa|grep mutt
mutt-1.4.1-5
root@box [~]# rpm -e mutt-1.4.1-5
root@box [~]# rpm -qa|grep openldap
openldap-2.1.22-8
root@box [~]# rpm -e openldap-2.1.22-8
error: Failed dependencies:
liblber.so.2 is needed by (installed) libuser-0.51.7-2
liblber.so.2 is needed by (installed) autofs-3.1.7-42
liblber.so.2 is needed by (installed) mod_perl-1.99_12-2
liblber.so.2 is needed by (installed) apr-util-0.9.4-2
liblber.so.2 is needed by (installed) php-ldap-4.3.3-6
liblber.so.2 is needed by (installed) nss_ldap-207-6
libldap.so.2 is needed by (installed) libuser-0.51.7-2
libldap.so.2 is needed by (installed) autofs-3.1.7-42
libldap.so.2 is needed by (installed) gnupg-1.2.3-2
libldap.so.2 is needed by (installed) mod_perl-1.99_12-2
libldap.so.2 is needed by (installed) apr-util-0.9.4-2
libldap.so.2 is needed by (installed) php-ldap-4.3.3-6
libldap.so.2 is needed by (installed) nss_ldap-207-6
@@@@@@@@@@@@@@@@@@@@@@@@@@@@@

As you can see if I want to remove cyrus-sasl its not possible cause openldap needs it. If I remove openldap, its not possible cause many modules depend on it including autofs :(

Is cyrus-sasl an important module? I thought about doing a force upgrade of cyrus-sasl to the centos version. Is it worth a try or should I wait for Steve? :(

How about openssh? I do have telnet activated in case something happens..

Originally posted by Mike_R
Is CentOS any better than Redhat 9 performance wise ?? I mean the loads and speed..

CentOS3.1/RHEL3 are built off of redhat 9 technology so you would probably see similar performance. No complaints here!

Originally posted by TheSpidre
As you can see if I want to remove cyrus-sasl its not possible cause openldap needs it. If I remove openldap, its not possible cause many modules depend on it including autofs :(

Is cyrus-sasl an important module? I thought about doing a force upgrade of cyrus-sasl to the centos version. Is it worth a try or should I wait for Steve? :(

How about openssh? I do have telnet activated in case something happens..

Did a --force --nodeps for the openssh and cyrus rpms (Centos version) which solved the dependencies.

Unfortunately, when I ran "yum upgrade", it did not update any packages except the following ones:

comps-extras-9.0.4-1.centos.1
redhat-config-packages-1.2.9.2-1.centos.0
MAKEDEV-3.3.12-1.centos.0
up2date-4.2.33-1.centos.3.1
dev-3.3.12-1.centos.0
basesystem-8.0-2.centos.0

I think it is because it cant find newer packages (centos is older).

In other words, it doesnt seem to be possible to do FC1 to centos :(

I am currently reviewing the installed rpms on a rh9 box (cpanel) that I plan to upgrade to centos, and I was wondering if anyone could confirm wether or not the follow rpms are needed.

XFree86

XFree86
XFree86-100dpi-fonts
XFree86-75dpi-fonts
XFree86-Mesa-libGL
XFree86-Mesa-libGLU
XFree86-base-fonts
XFree86-devel
XFree86-font-utils
XFree86-libs
XFree86-libs-data
XFree86-tools
XFree86-truetype-fonts
XFree86-twm
XFree86-xauth
XFree86-xdm
XFree86-xfs

Gnome
gnome-libs
gnome-mime-data
gnome-vfs2

The only ones you'll need out of the above are:
XFree86-Mesa-libGL-4.3.0-2.90.55
XFree86-devel-4.3.0-2.90.55
XFree86-libs-data-4.3.0-2.90.55
XFree86-font-utils-4.3.0-2.90.55
XFree86-libs-4.3.0-2.90.55

Some of the above may not be required, depending on what you have installed, however you need no more than what I've pasted above. The rest of those rpms you have listed can and should be removed.

Having a problem, i have done the upgrade, with no errors, and when i now run yum update, there are no packages to update.....

I reboot the server, it comes back up with the new kernel no probs but when i run

cat /etc/*release*


it comes back as

Red Hat Linux release 9 (Shrike)

:(

Any ideas???

Originally posted by IWebSolution
Having a problem, i have done the upgrade, with no errors, and when i now run yum update, there are no packages to update.....

I reboot the server, it comes back up with the new kernel no probs but when i run

cat /etc/*release*


it comes back as



:(

Any ideas???

# rpm -e redhat-release
# rpm -ivh centos-release

i have the same prob as IWebSolution
The upgrade worked without any errors

$ cat /etc/*release*
Red Hat Linux release 9 (Shrike)


on:


# rpm -e redhat-release
# rpm -ivh centos-release

i get:

$ rpm -e redhat-release
error: Failed dependencies:
/etc/redhat-release is needed by (installed) autofs-3.1.7-36
/etc/redhat-release is needed by (installed) initscripts-7.14-1
$ rpm -ivh centos-release
error: open of centos-release failed: No such file or directory
$


the Kernel it installed is: 2.4.20-31.9

Excellent tutorial, thanks (yeah, I use this stuff occasionally;))

if you run into problems with exim (I did) and Cpanel after doing this, run /scripts/exim4 which will reinstall exim and those problems should dissapear

Again, excellent tutorial, been meaning to test it out. I ran into a few problems, but those were easily overcome by skill ;)

Thanks!

Just used this tutorial as the base to upgrade From RH 9 to CentOS 3.4.

It works like charm.

Only had to add the new kernel to my grub.conf manually, rebooted and:

Linux server.host.here.com 2.4.21-27.0.1.EL #1 Fri Dec 24 02:04:03 GMT 2004 i686 i686 i386 GNU/Linux

This was in a cpanel box.

Exim refused to work, this was fixed by updating cpanel to latest version and now everything is working good.

Thanks a lot!

You'll need to do a bit more than upgrade exim to get this to work perfectly. Various other problems will be seen. I'd suggest a cpanel reinstall (/scripts/upcp --force) to reinitialize the perl modules that have been screwed up in the upgrade, a exim reinstall (which you've done), and an apache recompile. This should clear up everything.

Apache needs to be reinstalled because of modules that were compiled against the old modules in rh9. I believe php won't work unless you do this (at least it hasn't for me, yet).

Thanks a lot for the pointers, ill reinstall cpanel and rebuild apache just to make sure.

Php is working fine on all the sites on the server.

The only issue im seeing is that GD wont install when i go to WHM Update server software, lets see if the cpanel reinstall solves that.

Well, the good news: GD finally installed after running: "/scripts/upcp --force"
The bad news is that File-Scan-ClamAV-1.8 wont.
But i have clamAv disabled due to high resources usage.

Removing previously used /home/.cpan/build/File-Scan-ClamAV-1.8

CPAN.pm: Going to build C/CF/CFABER/File-Scan-ClamAV-1.8.tar.gz

/usr/bin/make -- NOT OK
Running make install
make had returned bad status, install seems impossible
perlmod--Install done
GD installed ok

If anyone else stumbles across this here are the EXACT steps to do this. I have done it one freshly loaded RedHat 9 and one Productions RedHat 9 box that has been used for a couple of months. No problems with either yet. If any arise I will post.

At the shell command prompt

rpm -vv --rebuilddb

rpm --import http://beta.centos.org/centos/3.4/os/i386/RPM-GPG-KEY-CentOS-3

rpm -U http://beta.centos.org/centos/3.4/os/i386/RedHat/RPMS/centos-release-3-4.2.i386.rpm

yum upgrade

yum update

/scripts/upcp --force

/scripts/exim4

cat /etc/*release*
(this should return CentOS release 3.4 (final)
If it does

shutdown -rf now

Tada... when it comes back you are upgraded.

Originally posted by kaveyhosting.com
If anyone else stumbles across this here are the EXACT steps to do this. I have done it one freshly loaded RedHat 9 and one Productions RedHat 9 box that has been used for a couple of months. No problems with either yet. If any arise I will post.

SNIP
You forgot the part about modifying your yum.conf so that your yum know's where to pick up those updates ;)

No this was done with a standard yum config file.

Yesterday tried to upgrade from RedHat9 to CentOS4.1
Everything was almost easy!
First upgraded to CentOS3,
Then simple to CentOS4

Im getting the following:


.Package php needs libgd.so.1.8, this is not available.
Package php needs libcurl.so.2, this is not available.
Package php needs libpspell-modules.so.1, this is not available.
Package php needs libpspell.so.4, this is not available.
Package httpd needs libdb-4.0.so, this is not available.
Package perl needs libdb-4.0.so, this is not available.
Package php needs libdb-4.0.so, this is not available.
Package perl-DB_File needs libdb-4.0.so, this is not available.



any ideas?

If you're using CPanel, remove those rpms (http, php, perl, etc) and their dependencies. You can reinstall perl using CPanel's setup, which is always better anyways.

Originally posted by linux-tech
If you're using CPanel, remove those rpms (http, php, perl, etc) and their dependencies. You can reinstall perl using CPanel's setup, which is always better anyways.

nope, this is a box with no control panel, i just want to make sure once i upgrade it i can have everything back to normal and working well

so you would reccomend removing those then once centos is installed just reinstalling them?

Yeah, if you can handle the sites being down for a few hours while the upgrade takes place ;)

Originally posted by linux-tech
Yeah, if you can handle the sites being down for a few hours while the upgrade takes place ;)

hmm i can stand that on this server, since its just a test server, but any way around this downtime on a production server? or just have to suck it up and do this at a time that my customers wont be angry at lol:)

also another question, can i do the straight upgrade from red hat 9 to centOS 4?
Or should i do redhat 9 -> centos 3.1 -> centos 4?
how would i do the upgrade from 3.1 to 4? any good guides?

thanks

Upgrading to 4.1 is tricky.. There are a few guides on the internet. But do to all the changes in 4 I DO NOT TRUST THEM. Theres just to much that can go wrong. You are changing from devfs to udev..