SAN attached device's SFP will be replaced with device name change?

[f. halili]

Hello,

I have a tape library and the SFP of the tape library where it is connected wll be replaced.

I know I have to change the zoning to account for the WWN change. My question is how will the OS ( HPUX ) behave with this change? Will the device names ( /dev/rmt/xxx ) change?

Thanks,
f. halili

[Torsten.]

I don't know what library you have, but normally the device keeps the WWN where the SFP is plugged in. The SFP itself doesn't have the WWN. Hence no change.

[Mel Burslan]

as long as the blade/port number on the FO switch stays the same and you do not move the FO cables on the server end, there is no reason for the device paths to change. But stranger things happened in the past. Hence do not take this as gospel word. Your milage may vary as usual.

[f. halili]

I guess as what both of you said. No change is needed. Even the zones won't change as the zone have the devices WWNs. There's no entry in the zone on the WWN of the port. As long as HOST's and TAPE LIBRARY's FC are the same, there should be no change in the device's name. .... I guess?

More comments are welcome.

Thanks,
f. halili

[Mel Burslan · Accepted solution]

If you know how the cXtYdZ number constructed, you can make the same deduction yourself.

cX: in the old days, when you talk about storage attached to any PA-RISC server, you were talking about SCSI interfaces. And on any given system, you had several of them. The number X designated the instance number of the SCSI interface and was dependent on which hardware slot this interface sat on. Numbering was hard to follow as it changed with the bus type, server model, almost from which way the wind blew. Today, fiber optic interfaces are usually attached to PCI/PCIx slots but still the X number given to them is related to which slot they are located at hardware wise.

tY: This is what-so-called the SCSI initiator number. Again in the old days of computing SCSI devices could only be daisy-chained upto 8 devices on on strand. One of these devices was the controller itself, which let you have 7 external SCSI devices attached to any given interface, SCSI initiator of each having to be a non-duplicated number in the range of 1 to 7. Then came fast wide SCSI and number range changed between 1-15, letting you attach 15 devices to any FW SCSI interface.

dZ: This is the device identifier for the lack of a better term. For those of us, who had misfortune of working with nikee arrays or jameca enclosures can relate to this better. You could access a bank of individual disks using one single SCSI initiator id. It is where this "Z" came to play a role. It designated which disk you were talking to through the single SCSI initiator. It ranged depending on the charateristics of the device you were talking to. Today, you can pretty much equate this to a LUN that your server is talking to using one path.

Having said all of this, regarding hpux history, today, a fiber optic interface is nothing more than a physical transport for an overlayed SCSI protocol and the numberting can be mapped about the same way.

Hope this can explain the concept a little better.

[Bill Hassell]

Ask your SAN administrator whether the tape drives are zone to a port rather than WWN names. Zoning to a port (N-port-ID) is ultra simple and often doe with tape drives because they don't seem to change much. And if you replace a drive, the new WWN is not important. The system sees the device at the port level. WWN changes means that you need to associate a device file (disk or tape) with a given hardware path. That is where fcmsutil uses the replace_dsk option to remap a changed device (replace_dsk really means replace_something).

[f. halili]

thanks all.