rp8400; 1 nPar; using 2 cells; question on drive availability

[Ed Lawrence_1]

Our lab has an rp8400 with 2 cell boards, 2 DVD-drives, and four hard drives. If you are familar with the rp8400, you know how the DVD's and fixed drives are arranged in two rows on the front.

When we set up 2 nPars, each partition has access to a DVD, 2 fixed drives, and 8 IO slots. I installed HP-UX 11i on a drive in each partition. The OS seems to work fine.
(Please note: I did a basic OS install on each drive using an Ignite server. I did not install any patches or updates).

I did the following, and I don't understand "why" it is working as it does.

Goal: Set up 1 nPar with 2 cells so the boot drive has access to all internal drives.

1. I removed the second nPar.
2. I re-created the genesis partition.
3. I booted to the OS and I added the 2nd cell to the partition using parmgr commands.
4. Reboot for reconfiguration.
5. Interrupt boot, do a SEA ALL. All four fixed drives appear.
6. Boot to OS. 'ioscan -ba' shows all 16 IO slots.
7. "ioscan -fnC disk" shows only 2 fixed drives and 1 DVD.
8. ioscan -fn shows several "PCI SCSI UNKNOWN" devices.
9. I rebooted and interrupted the boot. SEA ALL shows all the drives. I booted manually to the boot drive that did not appear in the ioscan previously.
10. Logged on and did "ioscan -fnC disk" and this drive and OS display all the drives.

So, the boot drive that originally was installed when it was part of partition 0 cannot see all the drives.

The boot drive that originally was installed when it was part of partition 1 does see all the drives.

Why does one drive see all the devices, but the other not? Thanks.

[Ted Buis · Accepted solution]

Ed,

I am not sure, but it may be possible that since your kernel was built when there was only one cell, it isn't seeing the second core I/O board that supports second pair of disks. Does ioscan see the second core I/O board as claimed? If the data on the disks isn't important to you, I would simply reinstall the kernel from DVDs.

Ted

[Ed Lawrence_1]

Thank you Ted,
Working on this during the weekend, we discovered that was the easiest solution (as it is a lab system). We also dicovered we can add a whole bunch of drivers through SAM(still trying to isolate which ones besides 'sdisk' make a difference.

It appears the 'best practice' is to build the partition then install the OS. While the approach I used in the lab has worked at times, it is not the best practice.

[Ted Buis]

If the second core I/O board is unclaimed then you likely need to add some driver to the kernel. From "sam" you might be able to figure out what is missing add the right one, but I don't know exactly which one it would be.

[Ted Buis]

Don't use the -n option on I/O scan, as it will only list what is in the /dev directory. Do an "ioscan" without any parameters and see if you see the device. If so, the do an "insf -e" to install the device special files.

[Ted Buis]

Sorry, I didn't see your reply until after I had and sent my other thoughts, but I think the -n option is why you couldn't see them initially. Reinstalling the OS certainly has the desired result of pulling in all the needed drivers.

[Ed Lawrence_1]

Thanks! I appreciate the inout and advice.
e.g. I didn't know the "-n" would limit the ioscan output.
Sorry I closed the thread so fast. Your answer confirmed what I suspected. Plus, if I don't assign points and close threads quickly, I often forget to do it for weeks.

[Ted Buis]

You assigned enough points. It doesn't make sense to give more than the max for a single question to my thinking.

Good luck,

ted