RH / Kickstart issues

[Mark Koenig]

1) I have a cluster of 128 ia32 nodes, using network loads via pxeboot and kickstart.
When I load RH WS3.0 U1, this all works fine and my nodes load as expected without problems(see question below).
If I attempt to load RHWS3.0 U2 or U3, none of this works and I get the standard install menu choices on the console of each node (i.e. none of my Kickstart choices are loaded). Did something change from U1 to U2/U3 that prevents this sort of auto load from happening?

2) When reloading my ia32 nodes after the root disk has been zeroed out, the kickstart installation process asks me if I want to continue, possibly destroying the data on the disk (even though it's zero'd out). I presume this is because the partition table has been destroyed? Is there a way to "automate" the answering of this question ("Yes") so that all of my 128 nodes will auto load without intervention for this type of reload??

If anyone can assist it would be much aprreciated!

Mark

[John Pretti]

Mark,

You can accomplish this by adding the below in your ks.cfg:

clearpart --all

This will remove all existing disk partitions.

Hope this helps.

Cheers,
John

[Jerome Henry · Accepted solution]

Hi,

It seems that some options don't work very well (like --bootproto static), whereas new kickstart package isn't really supposed to change anything...
A good way round is to add, in your command section :
autostep
or :
interactive
They will both show you ks option and go to next step (interactive waits for your ok, autostep goes by itself) : you should be able to see where it blocks...

hth

J

[Mark Koenig]

Thanks much!!

It appears several of the kickstart parameters "slightly" changed from U1 to U2.

I'm loading U3 now. :)

Any thoughts on how to get around the second part of my question? We'd like to zero out the disk(s) each a customer is done with machine before we reload the OS, but having to answer the question slows the process down.

Again, many thanks!
Mark

[Stuart Browne]

So you not only want to destroy any existing partitions, but want to zero-it-out first?

Or you're saying it's already been zero'd out, and even though you've got 'clearpart --all', it's prompting you with the confirmation dialog?

Possibly in the zeroing-out phase, make a dummy partition with nothing on it or even formatted. it should not reach that area of logic.

[Mark Koenig]

Yes, we want to zero the disk to assure the customer that it's all clear, then when the node is power cycled, it begins to reload over the network, at this point is when the kickstart process asks if it's ok to destroy the disk, which means I need operator intervention on the console...multiplied by 128 :) (soon to be thousands).

I need to remove at least the /boot partition, otherwise the node will attempt to boot from disk, rather than do a reload. I've tried to skip the first few blocks on the disk, but it still hits the confirmation question.

Mark

[Jerome Henry]

Hi !

Couldn't you use in %pre section :
badblocks -w /dev/sda1 (assuming sda1 is your /boot ?)
Never tried it, but it would zero it, above erasing partitions.
What command do you use to zero it now ?

J

[Mark Koenig]

I need to "zero" the disk while it's booted for the customer...it's their last step before giving back the machine to us. Presently I "dd if=/dev/zero of={disk device(s)}

In this way, the customer is assured that it's all clean before we get it back.

I appreciate the feedback on this!
Mark

[Stuart Browne]

Try including the KS command:

zerombr yes

and see what it does.

If that still fails, build one up with interactivity, then view the 'anaconda-ks.cfg' and see what flags there are that you're not aleady using.

To give us a bit mroe of an idea of what's going on, try also attaching your 'ks.cfg' so we can have a look at what it's trying to do.

[Mark Koenig]

Aha!
That does it, thanks so much. I read right over that and didn't even see it.

Thanks to all for the help!
Mark