SAN problems fcmsutil stat errors

[PW HP-UX Support Team]

Hello.

I keep loosing luns on one path on one card. The lun's go from claimed to no_hdw, and are not accessable. A diskinfo gives me:

# diskinfo /dev/rdsk/c7t0d1
diskinfo: can't SIOC_INQUIRY /dev/rdsk/c7t0d1: No such device or address


An fcmsutil on the port gives me:
# fcmsutil /dev/fcd0 stat -s
Thu Apr 16 09:32:41 2009
Channel Statistics

Link Statistics from firmware ...
Link Failure count 3 Loss of Sync count 8
Loss of Signal count 8 Prim seq proto err count 0
Invalid Word count 3037044 Invalid CRC count 1


What is an "Invalid Word count"

The LED comes up solid yellow meaning it has synced at 4 Gb. Then it flashes or blinks like mad, then stops and will intermittently blink like mad again. But cannot access storage. I know the zoning is right and there are no errors on the switch port. Tried new cables, new ports.

Has anyone experienced this behaviour before?

Itanium, rx7640, 11i v3.

Tks

[Sivakumar MJ._1]

Check the SFP in both sides of the FIBER. U can check the HBA Card.

Let me know the connectivity of the FC from u r Server HBA, the switch type/model.

If brocade check with portshow and porterrshow...

[PW HP-UX Support Team]

Sorry, what is the SFP?

[PW HP-UX Support Team]

HBA is: HP AB379-60101 4Gb Dual Port PCI/PCI-X Fibre Channel Adapter

Brocade is Silkworm 5000 running OS 6.1.1c

The port on the switch has 1+ million invalid word count errors, whatever that is. Those errors occured when the HBA generated the same errors. The storage guy doesn't know what it means either.

The paths are spread over 2 switches and both switches have million+ invalid word count errors.

TIA

[Sivakumar MJ._1]

For SFP check the link

http://en.wikipedia.org/wiki/SFP_transceiver

[Sivakumar MJ._1]

For the error counts,

1.you can check with replacing the sfp

2.Fiber cable

3.GBIC Card

4.HBA card....

[skt_skt]

if you already covered fiber cable,switch port(GBIC) then you could change the HBA. In your case there is a DUAL port card so even you can swap the cable to the other port (if NOT used already) and update the masking with new WWN of the "other unused port"

[PW HP-UX Support Team]

Hello again.

For the record, I have changed the cables & ports 3 times, changed the card twice and the PCI slot once. Even had the tech replace the backplane on the rx7640. The invalid word count happens on every port on both switches when connected to the one hba in the one vPar.

The second vPar on that box, which is connected to the same switches/frabric works fine. AND if I use the that vPars cables/ports it still generates those errors when connected to the questionable hba.

Idea?