secure path & device

[Oftens]

My Environment is as following:
storage:EVA5000
host OS: HP-Ux B.11.11

I use "command view eva" to create & present a Vdisk to the Hpux host,
then I excute "ioscan -fnCdisk" & "spmgr display". The output is as attachment file.

My question are in the output with "spmgr display" of the new one vdisk
1.why the new one Vdisk's TGT/LUN is not 1/2 but 1/4
2.Device is not c12t1d2 but c12t1d4
3.Can this situation be changed?

thanks!

Best Regard,
Oftens

[Eric SAUBIGNAC]

Bonjour,

No eva 3000 / 5000 active/passive near me to make a test. Anyway ...

1) and 2) same question, no. No idea for the moment. Post an entire "ioscan -fn" and "spmgr display -u"

3) As far as I can remember, you can force tgt/lun. Try this :

- spmgr delete 6005-08B4-0001-0BF3-0002-5000-0022-0000
- ioscan
- insf -e
- spmgr add 6005-08B4-0001-0BF3-0002-5000-0022-0000 1 2
- ioscan
- insf -e

Regards

Eric

[Oftens]

Dear Eric,

I have tried your commands,but the situation remanin upchanged.

I found something stranger in the output of
"ioscan"
------------------------------------------
hsx 58 0/2/1/0.1.4.0.0.1.2 hsx CLAIMED DEVICE COMPAQ
HSV110 (C)COMPAQ
------------------------------------------
but I cannot see this Vdisk in the output of "ioscan -fnCdisk".

I attachment a file.

Please help to check it. Thanks!

Best Regard,

Oftens

[Amar_Joshi]

Oftens,

Have you checked on the LUN ID for the new drive from Command View EVA?

If you look for the host group and then presentations, you will find list of all the luns presented to this server with their LUN IDs. Looks like the LUN ID on the EVA is forcing your server to see the devices in that order.

LUN ID can't be changed without deleting the VDisk and recreate it again (**DATA LOSS**).

[Uwe Zessin]

Unfortunately, the term "LUN ID" is a bit ambiguous, but I think that you are not correct anyway.

The "LUN address" (1..255) can be changed by unpresenting the vdisk from a host and then representing it to a different address for that host.

The "LUN WWN" / "World Wide LUN name"
(6xxx-xxx-xxx-xxx-nnnn-nnnn-nnnn-nnnn), which by default is the same as the internal UUID can be changed after unpresenting the vdisk from all hosts .

[Amar_Joshi]

Thanks Uwe,

Yes you are right, it can be changed with a downtime but not with data loss. I got it wrong between my own thoughts.

Thanks again and appreciate your correction.

[Eric SAUBIGNAC]

Bonsoir,

Sorry for late answer, but many work today.

You said :

>------------------------------------------
>hsx 58 0/2/1/0.1.4.0.0.1.2 hsx CLAIMED >DEVICE COMPAQ
>HSV110 (C)COMPAQ
------------------------------------------
>but I cannot see this Vdisk in the output of "ioscan -fnCdisk".

Yes you can. It is : /dev/dsk/c12t1d1.

Check with spmgr display, look for Device c12t1d1, and you will see 4 path : c8t1d2, c10t1d2, c9t1d2, c11t1d2 --> There is a direct relation between t1d2 from the 4 path and .1.2 at the end of the HSX path seen in ioscan.

But there is no direct relation between instanciation of spmgr device disk file and hardware path to the EVA LUN. What you must keep in mind is that spmgr does pseudo device numbering in the order in which it first sees a LUN.

For example, you can have a first LUN, number 26, seen as HSX path .......0.3.2 (3x8 + 2) and spmgr will associate pseudo device /dev/dsk/c12t0d0 (c8t3d2, c10t3d2, c9t3d2, c11t3d2).

Then create later a second LUN, number 13, seen as HSX path .......0.1.5 (1x8 + 5) and spmgr will associate pseudo device /dev/dsk/c12t0d1 (c8t1d5, c10t1d5, c9t1d5, c11t1d5).

With spmgr and active/passive array, the harder work is to identify correctly spmgr pseudo disk devices and corresponding LUNs in EVA. You will find in attachement an excel summary of your spmgr devices, built by reconciling output from ioscan -fn and spmgr display.

In a summary, I see no problem if there is no continuity in spmgr pseudo devices numbering : LUN 1 to 10 from EVA are c12t0d0 to c12t1d1, but LUN 11 and 12 are c12t1d3 and c12t1d4, not c12t1d2 and c12t1d3, that's life !

Anyway, it is a bit surprising that the new LUN was not associated with c12t1d2 since this position seems to be free. I guess that if you "spmgr delete 6005-08B4-0001-0BF3-0002-5000-0022-0000", then unpresent this VDISK from EVA to the HP-UX box, then reboot the HP-UX box, then present back this VDISK with same LUN number (12), ioscan, insf -e, this LUN will be associated with c12t1d2 as you want.

But do you really need to do that ? When you know that spmgr does pseudo device numbering according to the order in which luns appears, is it really important ? What is, in fact, important is to know how to identify correctly spmgr pseudo devices and to maintain a table of spmgr devices.

Regards

Eric

[Oftens]

Dear Eric,

I have two nodes with ServiceGuard.

For consistency, I want the new LUN associate with the same device c12t1d2 in both nodes.

I will try you suggestion to see if this LUN will be associated with c12t1d2.

Thanks!

Best Regard,

Oftens

[Oftens]

Dear Eric,

It is not work with your procedures.

I have tried the following procedure:

Step1:shutdown all application and unmount all filesystem related to LUN11(device:C12t1d3)

Step2:vgexport to create mapfile to backup all vg

Step3:vgexport -v vgname(which is related to LUN11)and vgchange -a n vgname

Step4:spmgr delete c12t1d3(LUN11) &c12t1d4(LUN12)& ioscan & insf -e

Step5:unpresent these two LUN in EVA command_view

Step6.# spmgr display -s
==> I got a stale LUN
Undiscovered Units Listing
= = = = = = = = = = = = = = = = = = = = =
Storage: 5000-1FE1-5001-5490
LUN ID: 6005-08B4-0001-0BF3-0002-4000-0043-0000

Step7:check 6005-08B4-0001-0BF3-0002-4000-0043-0000 is not in the output of"spmgr display"

Step8:# spmgr clean -d 6005-08B4-0001-0BF3-0002-4000-0043-0000

Step9:# spmgr display -s
==>to check there are no undiscovered units.

Step10: reboot

Step11: present original LUN12 as LUN11 back to the host and run ioscan& insf -e. It got the next position TGT/LUN=1/2,device=c12t1d2.
(This is what I want)

Step12: present original LUN11 as LUN12 back to the host and run ioscan& insf -e. It got the next position TGT/LUN=1/3,device=c12t1d3.

Step13: restore vg with vgimport -v vgmapfile & vgchange -a y vgname & mount all needed filesystems.

Step14: start all needed application and the system is back to normal operation.


Thanks!

[Oftens]

I have posted the solution.