Fibre Channel Driver detected a parse error in FLOGI/PLOGI

[Francis Noël]

Hello wizards,

I am new to the Unix world and I need some advice.

I have a server here complaining about some FLOGI/PLOGI parse errors. The server is an L class HP9000 with 2 Tachyon adapters. The SAN is an IBM Shark. Here is a snip from syslog.

[snip]
0/12/0/0: 'World-wide name' (unique identifier) for device at nport ID 0x201a00 has
changed. If the device has been replaced intentionally, please use the
fcmsutil(1M) command's replace_dsk option to allow the new device to be used.
0/12/0/0: Fibre Channel Driver detected a parse error in the FLOGI/PLOGI response
returned by nport ID 0x201a00. FLOGI/PLOGI Fail Code = 0x6.
[/snip]

The message repeats for 0/10/0/0, the second fibre adapter. nport remains 0x201a00.

It is my understanding that this error will show up when hardware gets replaced, i.e. replace a failed disk or fiber adapter , and that the affected device will be unavailable until fcmsutil replace_dsk is issued by the admin.

Weirdness begins here as nothing has been replaced in the loop and/or Shark and every Oracle database on the server is chugging along fine.

What did happen, progressively, is that some LVM volume groups on the server have been reduced and the corresponding LUNs on the shark have been removed from the server's available pool of storage. Some other VGs have then been extended later on and I believe that the same storage areas have been reassigned to the server, maybe under a slighly different configuration.

The Shark keeps track of storage units ( or LUNs ) with " serial numbers ", which are apparently reusable.

I am still unclear on much of the Shark's terminology and concepts so I might have left something out here, dont hesitate to ask questions.

I think a reboot would have the server rebuild its WWN lists but this is a 24/7 machine, so not an option.

Here are my questions :
Is is possible that the FLOGI/PLOGI errors stem from a LUN with serial number X being removed from the server and then re-added with the same serial number BUT with a different size ?

How can I find out exactly what is at nport 0x201a00 ?

Ultimately I'd like to use fcmsutil replace_dsk to fix this, is it safe ? How do I do it ?

Thank you for taking the time to read through this lenghty post.

[Francis Noël]

Edit :

I spoke with the Shark people ( nice folks )and they tell me that once a LUN is created, it is permanently assigned its serial number and its size remains the same.

My question remains valid though. Can this error be caused by a LUN being assigned, used, vgreduced, removed from the storage pool ( fibre driver sees it as inactive/disconnected ) and then reassigned and reavtivated ?

[AwadheshPandey]

go:
http://docs.hp.com/en/diag/ems/dm_TL_adapter.htm
http://forums1.itrc.hp.com/service/forums/questionanswer.do?threadId=363026&admit=-682735245+1129794491640+28353475
http://forums1.itrc.hp.com/service/forums/questionanswer.do?threadId=860891

Awadhesh

[Francis Noël]

Thanks for your reply Awadesh

I'd already read all of those, still no way to know what is at nport 0x201a00.

I went ahead and did
fcmsutil /dev/fd0 replace_dsk 0x201a00
fcmsutil /dev/fd2 replace_dsk 0x201a00

The server has stopped complaining but no new devices show up after an ioscan, insf -e, cfgvpath -r and datapath query device.

I dont feel comfortable issuing commands " blindly " like that so if anyone has a way to determine what is behind a nport I'd appreciate it.

I know that nport 0x201a00 is not one of fc controllers on the server so it has to be either a switch in the loop, a port on the SAN or a LUN ( ? ).

[Denver Osborn · Accepted solution]

the nport id is made up from the switch domain id, area and port

0x201a00

domain 20 (hex)
area 1a (hex)
port 00 (al_pa for fabric attached is 00)

the device you could be looking for is connected to port 10 of switch w/ domain id 32. I think the port number could be off, depending on the switch pid format...

To get additional info (wwn, etc), you can also use fcmsutil.

fcmsutil /dev/td0 get remote 0x201a00

hope this helps,
-denver

[Jeff Schussele]

Hi,

I suspect something may have changed on a SAN switch in the path.

Rgds,
Jeff

[Francis Noël]

Hello again

Thank you for your replies

Denver you hit the mark with fcmsutil /dev/td0 get remote 0x201a00.

With this command I was able to obtain the WWN associated with 0x201a00 : it is the WWN of the N_Port of the fc card on another server.

I need to speak with some people here to figure out what happened to that card/server.

Stay tuned :)