Where to permanently deactivate FCMS HBA?

[Ralph Grothe]

Hi there,

I've got one box which has a redundant, and unused 3rd Tachyon HBA, which however is generating FLOGI parse errors that trigger unnecessary alerts of my Nagios monitors of this server.

# ioscan -knfdtd
Class I H/W Path Driver S/W State H/W Type Description
=================================================================
fc 0 0/4/0/0 td CLAIMED INTERFACE HP Tachyon TL/TS Fibre Channel Mass Storage Adap
ter
/dev/td0
fc 1 1/4/0/0 td CLAIMED INTERFACE HP Tachyon XL2 Fibre Channel Mass Storage Adapte
r
/dev/td1
fc 2 1/10/0/0 td CLAIMED INTERFACE HP Tachyon TL/TS Fibre Channel Mass Storage Adap
ter
/dev/td2


1/10/0/0 has no storage attached

# ioscan -knfH1/10/0/0
Class I H/W Path Driver S/W State H/W Type Description
===================================================================
fc 2 1/10/0/0 td CLAIMED INTERFACE HP Tachyon TL/TS Fibre Channel Mass Storage Ad
apter
/dev/td2
fcp 2 1/10/0/0.8 fcp CLAIMED INTERFACE FCP Protocol Adapter


but clutters syslog with these

# grep 1/10/0/0.\*parse\ error /var/adm/syslog/syslog.log|tail -1
Dec 1 09:09:19 neptun vmunix: 1/10/0/0: Fibre Channel Driver detected a parse error in the FLOGI/PLO
GI response

# grep -c 1/10/0/0.\*parse\ error /var/adm/syslog/syslog.log
33


The fcmsutil's stats reflect these parse errors

# fcmsutil /dev/td2 stat|grep -i -e flogis\ sent -e parse\ errors
FLOGIs sent 2145854
FLOGI Retries skipped due to Parse Errors 2145160
FTO OCMs for FLOGIs sent 2145131


as compared to the HBA with the primary path storage attached where there are none


# fcmsutil /dev/td1 stat|grep -i -e flogis\ sent -e parse\ errors
FLOGIs sent 2
FLOGI Retries skipped due to Parse Errors 0
FTO OCMs for FLOGIs sent 0


Of course I can instantly end this misery


# fcmsutil /dev/td2 -f disable

Adapter is Disabled

# fcmsutil /dev/td2|grep -i state
Driver state = OFFLINE


but this setting wouldn't survive a reboot I assume.

Searching for an appropiate config file in the most obvious place I came accross this one



# cat /etc/rc.config.d/hpfcmsconf
#####################################################################
# @(#) hpfcmsconf $Revision: 1.1.119.1 $ $Date: 97/07/18 12:44:41 $
# hpfcmsconf: Contains configuration values for HP Fibre
# Channel Mass Storage interfaces
#
# FCMS_START=[0,1] Set this flag to 0 to NOT create device files
# HP_FCMS_INTERFACE_NAME Name of interface (fcms0, fcms1...)
#
#
#####################################################################

FCMS_START=1
HP_FCMS_INTERFACE_NAME[0]=

# End of hpfcmsconf configuration file


but one thing that puzzles me is that the array elements in HP_FCMS_INTERFACE_NAME seem nowhere in an init script to be referrenced again


# grep -l HP_FCMS_INTERFACE_NAME /sbin/init.d/*


I feel the strong urge to populate this array like such

HP_FCMS_INTERFACE_NAME[0]=/dev/td0
HP_FCMS_INTERFACE_NAME[1]=/dev/td1


and hoping thus /dev/td2 will never get activated on boot up,
but I fear this will be pretty useless as long as this array is nowhere ever accessed again.

Regards
Ralph

[Denver Osborn]

Hey Ralph...

did you test out the above config and did it keep the port disabled after a reboot? Just curious as I'm looking to do the same for one port of a dual port A6826A HBA.

So far I haven't found any docs nor any hints in the install guide for the adapter.

-denver

[Greg Vaidman · Accepted solution]

Ralph,

I don't think making the changes you describe will have any effect, based on the contents of /sbin/init.d/hpfcms (timestamp in my version is: @(#)B.11.11_LR hpfcms $Revision: 1.1.119.1 $ $Date: 97/07/18 12:44:39 $)

Neither will changes to /etc/rc.config.d/fc_td_conf, as the script associated with that config file similarly doesn't reference the variables in the conf file! (Come on HP, what's up with that?)

You could try installing PHKL_33066 (FCMS patch) and PHSS_28509 (Tachyon TL driver patch) - if you're running 11.11, otherwise just search thru the patch DB for "fcms". Not sure if these will help, but it looks like the Tachyon driver patch at least updates the /sbin/init.d/fc_td script.

If it doesn't help, you could open a call to submit an enhacement request to these scripts.

Alternatively, you could write your own startup script, probably /sbin/rc2.d/S901hpfcms_disable that runs your disable command above.

[Steven E. Protter]

Shalom Ralph,

If the card doesn't need to be in the machine, why is it in the machine?

Perhaps disconnect the fiber cable from the back, reconnect it when you need to activate it. Its not like its ready for hot failover or anything.

Alternatively, these errors may be generated by the card detecting the disk array as lun 0, which in fact it is not. Assign a small lun 0 to the card and the problem might go away.

SEP

[S.Arunkumar]

Hi Ralph,

The fcmsutil what you mentioned clearly says the HBA is faulty.

# fcmsutil /dev/td2 stat|grep -i -e flogis\ sent -e parse\ errors
FLOGIs sent 2145854
FLOGI Retries skipped due to Parse Errors 2145160
FTO OCMs for FLOGIs sent 2145131

Due to which it is unable to login to the SAN Fabric.i.e flogi/plogi.

One more thing what you can cross check is the Fiber cable between this card and the Switch.

But in most of the cases I have seen the problem lies with the HBA.

Better you replace the faulty card.Else plug it out from the Host.Which will solve your problem.

Regards
S.Arunkumar.

[RAC_1]

You can certainly disable it with some scripting put at boottime. - Use rc scripts. the command to look at is rad. Particularyly following options.

# rad -N 1/0/14/0/0 (returns the slot ID)
# rad -a 0-1-1-6 (checks if other slots belong to the same power domain)
# rad -s 0-1-1-6 (suspends the operation of the slot)
# rad -o 0-1-1-6 (turns power off)
# rad -i 0-1-1-6 (turns power on)
# rad -r 0-1-1-6 (resumes the operation of the slot)

When putting a script, note than any changes to system-addition or deletion of devices may change it's hardware address and you may end pu putting off wrong hba.

[Ralph Grothe]

Dear Repliers,

sorry for my delayed feedback.
I spent last week at a RedHat Directory Server training course in Stuttgart
away from home and my mail box.

I agree that the best choice would be to disassemble the unused FCMSA altogether.
Though I think I could manage this myself
(in the end tinkering with one's PC doen't differ that much from opening up and unplugging a card from an enterprise server these days)
I wouldn't want to do this necessarily because the box has a valid support contract that I don't want to void by my unqualified tampering.

The suggested usage of various rad commands don't seem to apply here because the box is an N-class legacy iron with legacy 11.00 OS without any OLAR support to my knowing.

It's not a feat to put the fcsmutil disable call in an init script anyway.
I just have been curious whether there were some "official place" where to put that switch.

Many thanks for your contributions.

Ralph

[HP User-St. Louis-MO]

Hey Ralph,

For the problem I was having w/ an unused fcms port, hp gave us the option to disable the adapter using fcmsutil. As you can see this change won't survive a reboot...

to fix out problem we're opting to put loop backs on the unused ports.

from what I gather the /etc/rc.config.d/conf files were of no use to disable it.

-denver

[Ralph Grothe]

Hi Denver,

thanks for reminding me of the possibility of loopback gadgets.
Even the EMS alert seems to mention them

--8<--- snippet from EMS alert ----


2. If no fiber optic cable is attached to the Fibre
Channel Mass Storage Adapter, verify that the loopback
cable (p/n 1005-0624) is functional and securely
attached to the Fibre Channel Mass Storage Adapter.

-------------------------------->8---

I asked our SAN admin if she had such thingy
that she could spare.
Unfortunately not.
I for now disabled the cards and think about
physically withdrawing them on the next scheduled downtime.