disk hw path changed after fiber channel failure

[Sana Chaarani]

Hi,

We have several hp-ux servers running 11i
connected to a xp1024 enterprise storage.
The port in the fiber switch where one of our
servers is connected, failed. All disks viewed on this server using this fabric, went no_hw. If I want to change the connection from port 9 to port 8, for example, all hw paths of my disks will change , right ? How can I recover from this ?

Thanks for your help,

Sana

[Devender Khatana]

Hi,

You will have to use vgimport command on all the new devices achived after changing the port. This shall recreate the vginformation with the new device files and the path.

Morover if it is a mcdata switch, the port hardly goes bad & there is a SFP module which fits into the port, which generally goes bad. This module can be replaced easily among different ports.

HTH,
Devender

[D Block 2]

are you balancing HBA's on your server ? maybe auto-path ? So you are running for the moment fine ? If you. make certain you also run: vgcfgbackup on your vg's. and copy your /etc/lvmtab before you do any export/imports.

[Pradeep_29]

If you are using Emulex HBAs you can do persistent binding.

If you are using default Taychon cards, I am not sure you can do persistent binding or not.

But, as suggested, you would need to do vgimport if hw paths change. They may not change in all scenarios.

[Stuart Powell]

Sana,

Are the zones on the FC switch set up as port to port, or by aliases? I use a Brocade fabric switch and assigned the zones using the WWN addresses for each HBA. In my case I can move a FC connection from one port on the switch to another on the same switch with no change in device defintion.

Though I haven't worked in a port to port switch I would think that if you redefine the zone definition, maintaining the same zone, you shouldn't cause any path changes.

Stuart

[Marek Smejkal]

indeed your dev files changed changing the switch port, the best way is just replace SFP from unused port (or spare one) to the original switch port.
Marek

[Stuart Abramson]

If you have a High Availability configuration, i.e. - two HBAs connected to two switches connected to two disk ports, (which is the way your should have odne it) then you still have access over the other connection. You just lost one connection. Don't panic!

Don't shutdown and do a vgexport/vgimport. You can fix this on the fly.

Just fix the switch port, reconnect everything (live - while the application is running), do a discovery sequence (ioscan, insf -e) and vgextend the new disks back into the VG. (don't pvcreate them! They already have data on them!)

If you only had one path to the volumes, then you have to do vgexport/vgimport; but you should ALWAYS have two paths.

[Denis dube]

Bonjour Madame Chaarani,



Jâ Ai un vieux CV datant de 2001 qui vous appartiens et jâ aimerais entrer en contact avec vous pour des postes permanents chez des client de très haut niveau (salaire 100k + et situés à Montréal). SVP me contacter que lâ on en discute. Merci de me transmettre un CV à jour et votre nouveau numéro de téléphone si possible.



Cordialement,



Denis Dubé, CRHA

Conseiller en talents, Nexio Select

Nexio Technologies inc.

www.nexio.com

Bur.: 514-798-3707 # 2075

Cell. :514-574-3176

[Torsten.]

If you have (port-)zoning configured, every port number is essential.

Otherwise beside the internal HW-path (PCI-slot) and the LUN ID only the switch domain ID and the port on the switch where the storage is connected are important for the HW-path and the device file. The port where the server is connected is not.

So you can swap the port for the server, but not the port for the XP array.