Okay, here's my problem:

We have just upgraded the Redhat Linux kernel on some of our servers lately via Redhat Network Agent. The kernel and rdinit file was installed correctly as we could see them in the /boot directory and the bootloader .ini file as well.

However, after we rebooted the servers, they just went right back to the current kernel instead of the new kernel. Since our servers are operating on SCSI Raid 5 disk array currently and they need to boot up with a floppy boot disk each time, we make a new copy of boot disk with the following command:

mkbootdisk 2.4.9-31smp (the new kernel)

The problem is, we couldn't create the bootdisk this way! It returned the following error message:

Insert a disk in /dev/fd0. Any information on the disk will be lost.
Press <Enter> to continue or ^C to abort:
Formatting /dev/fd0... done.
Copying /boot/vmlinuz-2.4.9-31smp... done.
Creating initrd image...
gzip: stdout: No space left on device
done.
Setting up syslinux... cat: write error: No space left on device
done.

It looks to us that there is no enough space on the boot up disk to contain the initrd file! Which means we're stucked! Why did Redhat create a Ram file which is too big to fit on a floppy disk?? We don't understand what these people are doing. The only version that can boot correctly now is: 2.4.7-10smp & 2.4.9-34 (non-smp). We want to load 2.4.9-34smp because the old kernel 2.4.7-10smp is quite sluggish in performance. Any suggestion how we can work around this problem? Thanks.

Tom

Why do you need to boot from floppy? We use RAID5 and don't need to boot from Floppy, with a correctly configured system you shouldn't need to boot from a Floppy just to get your RAID card working.

You mean we can boot up the server with Raid 5?? I thought this is not possible, according to Redhat's documentation and several feedback I've received. It would be really great if we could boot up our server without having to use the boot up disk. Could you share how did you do that?

Our servers are configured with hardware Raid 5 array (3 HDD). I don't see any option to configure a non-Raid partition that we can use to boot up, either in the SCSI card BIOS or Redhat's Disk Druid configuration. We had been looking for a solution to this for some time now. We will really appreciate a lot if you know there is a way to work around it. Thank!

Tom

Hi,

Well to start you need to compile the RAID Card drivers in to your kernel so it knows how to talk to it, then it's a case of setting up your RAID array (You've done that part), then the OS should see the RAID array as one large disk with your partitions on it and you boot as you would normally, or that's the theory, I can't tell yuo for sure because I didn't set it up on our servers that use it.

Originally posted by itzme
Okay, here's my problem:

We have just upgraded the Redhat Linux kernel on some of our servers lately via Redhat Network Agent. The kernel and rdinit file was installed correctly as we could see them in the /boot directory and the bootloader .ini file as well.


"bootloader, .ini file"?


mkbootdisk 2.4.9-31smp (the new kernel)


Do you want/need SMP? Is this a dual+ CPU system?


The problem is, we couldn't create the bootdisk this way! It returned the following error message:

Insert a disk in /dev/fd0. Any information on the disk will be lost.
Press <Enter> to continue or ^C to abort:
Formatting /dev/fd0... done.
Copying /boot/vmlinuz-2.4.9-31smp... done.
Creating initrd image...
gzip: stdout: No space left on device
done.
Setting up syslinux... cat: write error: No space left on device
done.

It looks to us that there is no enough space on the boot up disk to contain the initrd file!


You're writing the image as well as the other stuff it needs. The size of your kernel and anything else it needs depends on what you have. A floppy is a bad, slow way to boot.


The only version that can boot correctly now is: 2.4.7-10smp & 2.4.9-34 (non-smp). We want to load 2.4.9-34smp because the old kernel 2.4.7-10smp is quite sluggish in performance. Any suggestion how we can work around this problem? Thanks.

Tom

2.4.19 is the newest kernel, 2.4.18 has a good running record, why would you only upgrade to 2.4.9-x? In the future use lilo or grub or something and don't bother installing the boot image onto a floppy.

I think since hes upgrading via RPMs on redhat. Redhat released 2.4.9-34, they havent released rpms for anything else yet.

Hi Robert,

Originally posted by 2host.com


"bootloader, .ini file"?

I was referring to the .ini file of the bootloader, ie. grub.ini

Do you want/need SMP? Is this a dual+ CPU system?

Yes, our servers are installed with dual CPU and that's why I want to load the smp kernel.

You're writing the image as well as the other stuff it needs. The size of your kernel and anything else it needs depends on what you have. A floppy is a bad, slow way to boot.

I guess you could see in my earlier post that the only way for us to boot the servers currently is using the floppy boot up disk, because of the RAID 5 set up. Until we can find a work around solution, there's no other way to boot the server.

2.4.19 is the newest kernel, 2.4.18 has a good running record, why would you only upgrade to 2.4.9-x? In the future use lilo or grub or something and don't bother installing the boot image onto a floppy.

2.4.9-34 is the latest kernel rpm released by Redhat. I mentioned that we use Redhat Update Agent so this is the latest version available. We try not to build our own kernel because that will cause a number of dependencies problem.

The main issue right is again, the limitation of the boot up disk. If there is a way for us to boot the server directly from the HDD then this won't be an issue anymore.

Tom

Originally posted by itzme
Hi Robert,

I was referring to the .ini file of the bootloader, ie. grub.ini


I've not used grub much before and prefer lilo, so thanks for letting me know what you meant.


Yes, our servers are installed with dual CPU and that's why I want to load the smp kernel.


Of course. Sorry it seemed as though you were saying it wasn't currently using SMP, but I might have misread it.


I guess you could see in my earlier post that the only way for us to boot the servers currently is using the floppy boot up disk, because of the RAID 5 set up. Until we can find a work around solution, there's no other way to boot the server.


Create the same/similiar boot partitions on all drives. It doesn't take up much room at all and then no matter what drive is available first ,it will be able to boot off of it.


2.4.9-34 is the latest kernel rpm released by Redhat. I mentioned that we use Redhat Update Agent so this is the latest version available. We try not to build our own kernel because that will cause a number of dependencies problem.


I've never liked the idea of RPM kernels, at all. I'd not dare use them. I feel sdafer and in more control building from source (but that's just me). I'm not sure exactly what you are concerned about breaking, but I'd suggest you consider building from source. It's up to you though and if this works for you, then that's all that matters.


The main issue right is again, the limitation of the boot up disk. If there is a way for us to boot the server directly from the HDD then this won't be an issue anymore.
Tom

Again, (use lilo and) create similar /boot partitions on all the drives, is my suggestion. Always keep a floppy of the boot image handy anyway. Good luck.

i didnt read the bost in completion but ... offhand from reading the inital post - if you upgraded the RHN kernel and upon reboot it went back to current-kernel, then your booloader is simply not loading the new kern ?

open /etc/grub.conf

check what the default= line is set to and make sure it points to your kernel image.

Remember grub considers 0 a value space. So your first image definition in grub.conf, from the top would be 0 , and so on.

e.g:
default=0
timeout=10
splashimage=(hd0,0)/grub/splash.xpm.gz
title Red Hat Linux (2.4.19-smp)
root (hd0,0)
kernel /vmlinuz-2.4.19-smp ro root=/dev/hda3