Mirror disk on itanium-- unexpected results

[Edward Finneran]

Just to clarify a previous response:

- if you look at the lvlnboot output you attached, some of the information is key.

You'll notice the lvol1, 2 and 3 are specified as existing only on /dev/dsk/c2t1d0s2, or NOT the drive you are booted from.

The thing to remember is that in the BDRA (if they still call it that, Boot Data Reserved Area), when you boot from a drive, it looks in the area to determine where to find the filesystem containing the kernel, where primary swap is, etc. So, whether you boot from your primary drive or your secondary drive, the BDRA data is the same - namely, go look at lvol1 2 and 3 in vg00, wherever that is. So vg00, with both drives, gets activated, and hp-ux looks at lvol1, 2, 3 on c2t1d0s2, regardless of which boot drive you booted from.

So, it's not a question of some type of mirroring copying that new file you created on c2t1d0s2 to c2t0ds2 and you see it over there - you're actually looking at the data on c2t1d0s2 itself.

But they're right - try out DRD.

[Shea Sullivan]

Great catch, Edward.
This caused me a lot of trouble yesterday, as I forgot to setboot back to the primary disk. After I broke the mirrors, the system was still booting to the alternate disk and putting that kernel into memory. But the root filesystem (from lvlnboot) remained on the primary disk.

When I tried to update the kernel in this environment, the changes never took effect (using SAM, or building it manually with mk_kernel). Also, it broke dmesg, lanscan, and wreaked havoc on my server! I think its because it was still grabbing the old kernel from the boot area on the second disk. When I corrected setboot to the primary disk, the kernel in memory matched the kernel on the disk, and everything went back to normal.

In short, I learned to use caution with setboot and mirror-disk!