OK the updates and move to a new mirror have been completed. Whats new in AOOI 0.18:

CentOS 5.3 is supported for Network installs (1and1 people should still use Local installs.)
New mirrors have been established for the ISO's

None of this has been tested yet, so all feedback (successes and failures) would be very helpful.

I've come up with a simple solution for installing on 1&1 boxes where the current installer can't find the kickstart file.

I had come up with a solution last time for my Business Root I server (which of course, I completely forgot a few days ago when I ordered a new Root II server to take advantage of the 50% off to get a decent box that didn't hang every day)

The / partition is just big enough (even on the old root I server) to hold the necessary ISO's to install Centos 5.3 (the first 3 CD's) but even better, on the Root II server the default 1&1 partition setup has / as 9GB, big enough to throw the DVD iso on it.

Run the AOOI installer; just go with network install (it really doesn't matter since network or local both can't find the KS), Reboot. When the installer can't find the KS file, select cancel for interactive install, select Hard Disk, then /dev/sda1 or whatever it is that comes up as the first drive to look on, and bob's your uncle it will see the DVD iso and happily start to install.

Edit the network settings (and whatever else you want) during the install to use DCHP, and you'll be up and running with a nice clean 5.3 machine.

Oh and another tip if you're planning on running it on a newly ordered 1&1 box - if you're wondering why it takes so bloody long to login on SSH or resolve IP's when you're doing a wget, do yourself a favour and edit /etc/resolv.conf before you do anything - the 1&1 Centos image is semi-broken, it puts your server IP as the first nameserver in resolv.conf, which of course, isn't going to get you very far, so just delete that line with your IP or comment it out. (This issue is what actually got me to go with a fresh normal proper Centos install rather than just go with the 1&1 image to save the hassle of trying to find the problem.) To be fair I think the Centos 5.3 install did the same thing, so if its still slow you might want to check that first.

If you'll be installing a caching nameserver on the box (for instance bind which comes with Plesk) you might want put that entry back when the local caching nameserver is up and running.

Hi,

i'm trying to install centos 5 on an 1and1 server via AOOI and local install but it ends up in not finding the correct disk controller driver after choosing a normal installation without the kickstart file.
Of course the network installation doesn't work.
The output of lspci indicates that it's a NVIDIA SATA CONTROLLER.

When i try an installation with the local Centos 5.4.iso DVD it immediately crashes with
What can I do?

Thanks a lot
BruceLee

hmm, try passing in "nosmp" to the kernel when it boots up. That could be a buggy bios, which Ive absolutely seen kill an install before. Oh the other thing to try would be turning ACPI off with "acpi=off"

Ok, I will try this tomorrow. But doesn't nosmp deactivate multi core processors?
Then my multicore AMD Proc would be a single core one. Or do I miss something?

Thanks

BruceLee

Yes it does. You may have to do this to get Centos to install, and you could upgrade a newer kernel which works with your system and turn SMP back on.

Be preapred though that you may also be stuck without multiple cores if Linux doesnt support some element of your hardware completely. This can happen if the even the latest kernel can't sort out your hardware.

Hi,
i got to manage it by tweaking your aooi.sh script to use centos 5.4 iso disk 1-3 and locally storing the complete centos 5.4 DVD.
After reboot two things got my attention.
1st. The swap entries where missing in fstab so I ran this commands from you kickstart file:



df -h -l gives me:


fdisk -l


2nd error is:
fdisk /dev/sda1 e.g. ends up with this error:

writing doesn't help.
could it be this bug: https://bugzilla.redhat.com/show_bug.cgi?id=479224
How can i determine that i can ignore this message?
Any additional infos you need?
Anyhting else I should check now?
Can i sort of stress test the machine to make sure that i can run my customers on it?


Thanks a lot
BruceLee

Where did you end up storing the CentOS 5.4 DVD, in what was swap? The reason its been using the older version is because newer centos needed 3 or more CD's for a local install and there wasnt enough space. And of course this is all really because I hadn't sorted out what is breaking the network issues with 1and1.

I stored it in "/" . There was enough space for the three iso's and the dvd because it is 10gb big.
swap was default 1and1. if i remember correctly on sda2 and sdb2.
during the installation where no errors.
could this error be directly linked with the setup process?

booted into 1and1 serial console rescue mode and ran cfdisk:


could the free space be reason for it or the wrong partitioning before AOOI-ing the box?
if so, what is the recommended partition layout before AOOI-ing?
Thanks
BruceLee

AOOI will convert the swap partition to a regular file system, and then download the ISO's there. That way you can blow away all the other partitions and not remove the iso's at the same time. Once the install is complete, it converts those back into swap partitions (or it should rather... sometimes this doesn't work)

thanks. I will try it from scratch again with the 1and1 image. this time without the dvd. maybe it was not necessary.
i put the dvd on the disk just to make sure that i can try it with that if the cd option does not work.

i think i know why the network installation fails. I'm not 100% sure but i think it could be the reason:
1and1 is providing classless static routes via dhcp by using the dhcp option 121.
i think that anaconda just does not support this dhcp option 121 and therefore does not "ask" for the routes.
without the routes configured the traffic does not get through.
if anybody is able to patch anaconda this could solve the problem.
using %pre kickstart option and adding the routes is allready to late.
maybe tweaking initrd.img could solve it, too?!


Thanks
BruceLee

You know trying fedora 12 (with a more advanced anaconda) might be a good first step. If that handles their dhcp environment a backport would be the next option.

ok.i will try it this weekend with fedora12. i wonder if it helps.
would be great if it does and backporting is possible. this would simplify a lot.

1and1 is checking the partition table warning. nothing i can do about it.
tried dd-ing the whole disks and mbr too. Nothing helped.
Maybe it's nothing important and one of the few 1and1 special implementations

Back again with good news.

Installing Fedora12 (which has Anaconda 12.46) via Network-Installation runs fine in the 1and1 network environment
without any problems. This leads me to the conclusion that I might be right with the reason why centos installation makes so much trouble.
Hopefully there is someone who can backport the fedora12 anaconda to the centos one?!
That would be a real relief for 1and1 installations.
If someone is able to do this within the next time I would be happy to test it immediately.

The other error with the partition table is something the superb 1and1 technicians can't explain. They really told me that fdisk is not made for checking sda's but only md's. HAHA . So is send them the same error with the md's if that makes them happy.

Anyway. If i check the partition with fdisk /dev/sda and the option p it show up correct.
Other tools like parted, cfdisk etc. also give me the correct partition table.
After checking this behaviour on another 1and1 server and realizing that this server also has the error and
is running fine for years now I decided not paying any attention to this anymore.