Alternate Link - MirrorDisk/UX

[Alzhy]

I've 2 N-Classes with 2 18G internal disks, one I did not install, the other I installed and did bootdisk Mirroring using MirrorDisk/UX. I notice on the other server, the 2nd disk c2t6d0 is marked as an alternate disk to c1t6d0 instead of using MirrorDisk. Is this some kind of a special configuration.. part also of MirrorDisk/UX? It seems to me that they serve bot the same purpose and that is to ptotect the system disk..
root@ffhpxdb3 # pvdisplay /dev/dsk/c2t6d0
Device file path "/dev/dsk/c2t6d0" is an alternate path
to the Physical Volume. Using Primary Link "/dev/dsk/c1t6d0".
--- Physical volumes ---
PV Name /dev/dsk/c1t6d0
PV Name /dev/dsk/c2t6d0 Alternate Link
VG Name /dev/vg00
PV Status available
Allocatable yes
VGDA 2
Cur LV 8
PE Size (Mbytes) 4
Total PE 4340
Free PE 2504
Allocated PE 1836
Stale PE 0
IO Timeout (Seconds) default
Autoswitch On

[Michael Steele_2]

Please paste in or attach:

ioscan -fnC disk
lvlnboot -v
vgdisplay -v /dev/vg00

[James R. Ferguson]

Hi Nelson:

A mirrored disk and one with an alternate link are not the same. "Alternate" links are also known as "pv_links". The alternate device file is nothing more than a different *path* to the same physical device. A mirror is a discrete device unto itself, upon which data is replicated (mirrored).

Alternate links are designated "primary" and "secondary". You can actually have up to eight alternate links which are pathways to a device. LVM uses the primary or first path declared during 'vgcreate' to perform I/O. In the event that this path fails (e.g. the cable is pulled) then the "secondary" pv_link takes over to handle the I/O. Thus, pv_links provide high-availability.

Disk mirroring in HP-UX is implemented at the logical volume level. That is, logical volumes are mirrored, not physical disks nor volume groups. Mirroring means that the data is written to physically separate media. Thus if one mirror fails, the data is available from the other. Mirroring thus, too, provides high-availablity.

If you combine mirroring *and* alternate links, then you have increased your level of high-availablity. Not only do you have a second copy of your data, but you have multiple *paths* by which you can access the data should one path fail.

Regards!

...JRF...

[James R. Ferguson]

Hi (again) Nelson:

A good overview of disk and filesystem concepts and management can be found in the "Managing Systems and Workgroups: A Guide for HP-UX System Administrators" manual:

http://docs.hp.com/hpux/onlinedocs/B2355-90742/B2355-90742.html

Regards!

...JRF...

[Alzhy]

Okay.. but what does the above outputs tell us? c2t6d0 is a different physical disk and yet it is being used as an alternate link! If I do a pvdisplay on it it has the same usage as c1t6d0. I am pretty sure c1t6d0 and c2t6d0 are the N4000's 2 internal disks. Why is this so? I was thinking this is perhaps another way to use MirrorDisk/UX but the man pages seem to say as you're saying --- alternate links are other paths to the same physical disk -- ie. multi-ported disks!

[Mark Greene_1]

It's a function of the PV_Links component of MirrorDisk. If you have MirrorDisk, then you also have PV_Links. The alternate paths are established when you create the volume group. See the man page on vgcreate for more examples.

There is no specific documentation on PV_Links in the man pages.

HTH
mark

[Michael Steele_2]

You've got some problems here with two controller???s points to the same disk. One disk, two controllers. It???s a problem and you'll need an HP CE to correct it.

What's the history of this server? Was it delivered this way by HP? Pull out your sales order and manifest, if purchased from a channel partner then they'll have this also.

In short, you don't need an alternate or PV link or duplicate controller which is device c2t6d0. You need another disk and the c2 controller needs to attach to it.

Run diskinfo and try to pick up the serial number:

diskinfo -v /dev/rdsk/c1t6d0
diskinfo -v /dev/rdsk/c2t6d0

Good catch though. Should be a reward from management for finding this goof. :-)

When you've got another disk, same size and part number, also see 'diskinfo', follow the mirroring a boot disk procedure.

[Steven E. Protter]

What you have is failover capability Nelson.

If your controller c1 stops working, you won't have any negative impact, the system will use controller c2

Same disk, two pathes.

Its like driving from Chicago to Madison Wisconsin.

You can take the I-94 and go through Milwaukee or you can Take the I-90 and go through cheese land.

SEP

[Mark Greene_1]

"Why is this so? I was thinking this is perhaps another way to use MirrorDisk/UX but the man pages seem to say as you're saying --- alternate links are other paths to the same physical disk -- ie. multi-ported disks! "

Do lvlnboot -v to verify if they are really the same physical device, or not.

If you see something like this:

Physical Volumes belonging in Root Volume Group:
/dev/dsk/c0t6d0 (1/0/0/3/0.6.0) -- Boot Disk
/dev/dsk/c3t6d0 (1/0/1/0/0/1/1.6.0) -- Boot Disk


you've got two different disks, both set to be boot disks. The second one is NOT mirrored, but it will boot in the event that the first one fails by virtue of the alternate link being established.

HTH
mark

[Alzhy]

Thanks thus far to those that have responded. Again, I am sure we've 2 physical disks here - and this is an N-Class (rp7400). What I am not sure though is if the N-Class's 2 internal disks are multi-ported.

I am still wondering whay we have this config - c2t6d0 being marked on the root VG as an alternate link even if ioscan shows c2t6d0 as a separate physical disk.
Any more ideas?

[Mark Greene_1]

In addition to the lvlnboot, try this:

lifls -Cl /dev/dsk/c2t6d0

If that returns something like this:

volume ISL10 data size 7984 directory size 8 99/10/28 15:23:53
filename type start size implement created
===============================================================
ISL -12800 584 240 0 99/10/28 15:23:53
AUTO -12289 824 1 0 99/10/28 15:23:53
HPUX -12928 832 800 0 99/10/28 15:23:53
PAD -12290 1632 1700 0 99/10/28 15:23:54
LABEL BIN 3336 8 0 02/05/28 10:08:10

They the disk has the boot information on it. What might have been done is the disk setup to be a mirror of the primary boot disk, but instead of being activated as a mirror copy, it was setup as an alternate link. Whether by accident or design, I cannot say.

You big issue here is that you cannot see what is on the drive. What you can try, if you've the time and opportunity, is to boot from it if it's been setup to be a boot disk.

Power down your system. Remove the primary boot drive. Power-up the system. Interrupt the boot process and interact with the ISL. Scan for boot paths to confirm that the ISL sees the disk. If so, select that boot path and see if the system comes up. If it does, take a look at what is on the drive, and how old it is.

If that doesn't work, power the system back down, replace the boot drive, and bring it back up. You can then remove the alternage link and use the drive as a mirror disk.

HTH
mark

[Animesh Chakraborty]

"I've 2 N-Classes with 2 18G internal disks, one I did not install, the other I installed and did bootdisk Mirroring using MirrorDisk/UX."

Can you give us the mirroring steps you followed for boot disk mirrioring?

Thanks
Animesh

[Todd McDaniel_1]

All:

Please forgive me for digging up an old post...but I am curious for your answer.

Did you have a second disk or merely an alternate path along a different controller?

[Alzhy]

Subsequent reboots fixed this strange problem.. And yes indeed on this N-Class, the two are distinct physical disks..

I was relatively "RE-Acclimatizing" back then and getting familair with HP-9000 environments. Now that I've almost a year's worth of experience with more current HP OS, software and HW (and getting better at that)... I can deduce (theorize?) the following:

c2t6d0 probably was a DD copy of the primary OS some time back. hence, it was being treated as an alternate link or PV link...


But I could be wrong...

[G. Vrijhoeven]

Hi Nelson,

We once did this deleberately in a lab on a server that had no alternate link just to see if we could get it working. What we did was a dd from one disk to the other.
When you do a pvcreate LVM writes a "unique" number to the disk to id it. When you do a dd from the entire disk to a second disk this id is copied. When you do a vgextend LVM will think it is an alternate link since the numbers match.

HTH,

Gideon

[Alzhy]

Aha! so the theory is correct!
Thanks!