1820879 Members
3904 Online
109628 Solutions
New Discussion юеВ

Re: LVM problem

 
Bruno Bossier_1
Regular Advisor

LVM problem

I have a server connect with 2 fibre cards to a SAN. External storage array is a STK FLX240 (Engenio). Lun's configured in LVM volumegroups with PV Links.

Problem : last week-end, STK moved the fibre connections to new switches, added an array and adapted the zoning. Before this change, my lun's were seen through controller c6 and c4. Now after STK intervention, they are seen through controllers c17, c19, c22 and c25. I understand that I now see 4 controllers because they added an array, but I do not understand why the original controller id's have changed. I thought that the controller id's were dependent on the physical location of the fibre cards which did not change in my case ?

Any ideas ?

Bruno
10 REPLIES 10

Re: LVM problem

Bruno,

You don't state what type of FC switches you have, so I'm assuming Brocade (B class) or McData (M class).

The controller numbers assigned to storage on a SAN is in fact dependent on something called a FCID or Fibre Channel ID. This is a 6 digit hex number made up as follows on Brocade/McData swicthes:

0xDDAAPP

where DD = switch domain ID
where AA = switch Area (on Brocade/McData this is based on an offset of the physical port number)
where PP = switch port (NOT physical port) - unless your are using arbitrated loop somewhere, this is usually set to 0.

So on Brocade/McData SANs It doesn't matter where the FC HBAs for your HPUX servers are plugged in, but it *does* matter where the storage is plugged in. Move it to a different port and the area (AA) changes. Move it to a different swicth, and theres a good chance that the Domain ID (DD) will change. Change from fabric login, to FCAL using translative mode, and the port (PP) will change. My guess is that in your case they didn't make sure that the domain IDs were the same even if they did plug back into the same ports.

If you look carefully at your ioscan output you should be avle to pick out the decimal part of your FCID in the hardware path, after the hardware path for your HBA, and before the target/LUN of the storage ports.

HTH

Duncan

I am an HPE Employee
Accept or Kudo
Bruno Bossier_1
Regular Advisor

Re: LVM problem

Duncan,

Thanks for the clarification. The switches are Brocade 24000 switches.

I will have a look at the ioscan output as you suggested. I will also check with the STK person who did this intervention about these domain id's, area and ports, ...

Cheers,
Bruno
Denver Osborn
Honored Contributor

Re: LVM problem

Short answer is no, the FC HBA does not make up the H/W path for the LUN. It's a combo of the HBA, Switch domain, switch port... etc. So your hardware path changed with the array being moved to the new switch. This is why the OS sees them as a new set of disks.

If you have the output of ioscan prior to the move, compare the new LUN H/W path to that of the old ones.

Hope this helps,
-denver

Bruno Bossier_1
Regular Advisor

Re: LVM problem

I have been looking for this FCID with ioscan, but does not seem to succeed.

Can you give me an example ? Can I find this with fcmsutil ?

Bruno

Re: LVM problem

Bruno,

Maybe I didn't explain myself well enough - the FCID is part of the hardware path - here's an example disk hardware path from a Superdome:

7/0/10/0/0.1.24.0.0.8.0

- 7/0/10/0/0 - thats the FC HBA

- 1.24.0 - thats the FCID of the storage controller converted to hex thats a FCID of 0x011800

- 0.8.0 - thats the vbus/tgt/lun on the controller port

If I change my SAN configuration, its only the FCID that will change in the hardware path, but thats enough for HPUX to treat it as a different device.

HTH

Duncan

I am an HPE Employee
Accept or Kudo
marie-noelle jeanson_1
Trusted Contributor

Re: LVM problem

Bruno,

You can look at some FC hw path decoding information in the FC support guide:
http://docs.hp.com/en/J6374-90018/ch03s07.html?btnNext=next%A0%BB

Marie-Noelle
Devender Khatana
Honored Contributor

Re: LVM problem

Hi,

I assume you have recovered data by importing VG's.

Regards,
Devender
Impossible itself mentions "I m possible"
Bruno Bossier_1
Regular Advisor

Re: LVM problem

Yes, I recovered my volumegroups using vgimport.

Thanks to all for the explanations !

Cheers,
Bruno
Joseph Loo
Honored Contributor

Re: LVM problem

hi bruno,

i think u should look at the points u have assigned. 1 out of 152 (and counting) is not good.

show your appreciation to those who have assisted u in finding your answers.

regards.

what you do not see does not mean you should not believe
Devender Khatana
Honored Contributor

Re: LVM problem

Hi,

More that your thanks for your responses your points assigned to answers will be more helpful for forum.Your profile shows you never did that.

Regards,
Devender
Impossible itself mentions "I m possible"