topic Re: fcms Driver state in Operating System - HP-UX

fcms Driver state

Roger Baptiste — Thu, 21 Mar 2002 11:13:33 GMT

Hi,

I have couple of Fibre cards whose driver state shows up as RESETTING in the fcmsutil output. It happens randomly, but not frequently. I have the fcmsutil running in a loop checking the fibre card status and it shows up as RESETTING around 10 times a day. At that instant, ioscan shows up NO_HW corresponding to the disks connected to the specific Fibre card. But, the Syslog file shows no PV switching. The Vgdisplay works fine. it does not cause a problem.
Is it because it happens too fast and goes away?

This is on a N-class box/A5158A cards using fcT1 driver.

thanks
-raj

Re: fcms Driver state

melvyn burnard — Thu, 21 Mar 2002 11:27:33 GMT

This could be:
cable not seated correctly or bad.
bad hub/switch.
problem on disc FC card.

I would:
ensure cables are seated on both ends
get cables measured with light meter
get card sin the disc enclosures tested.

Is there a card that is NOT showing these problems?
If so, use it's cable to swap with a suspect cable and then monitor. This will prove cable or not.

Re: fcms Driver state

Tom Geudens — Thu, 21 Mar 2002 12:17:35 GMT

Hi,
mmm ... used to see this on our systems too ...

Do you have the latest fcms patch ? It used to be PHKL_23939 ... but there might be a more recent one.

Is the server connected directly to the storage ... or do you connect through a hub ? In the latter case the hub-statistics will show those problems. You might even catch it "online". Finding where the problem is will be difficult though.

In the former case I would put my money on a bad fibrecable.

Regards,
Tom Geudens

Re: fcms Driver state

Roger Baptiste — Thu, 21 Mar 2002 12:56:26 GMT

hi,

The cards are directly connected to Xp array. There are four fibre connections to the disks and there are four tachlitefibre(td) to the tape drives. All direct connections.

I am not sure from when this problem of resetting message started occuring. I started monitoring it closely couple of days back, after we had some performance issues with the backup running very slow on the box. This started happening after a couple of new fibre cards were added for future usage. Those cards were then removed, due to this problem. But still i see random reset messages.

thanks
raj

Re: fcms Driver state

Roger Baptiste — Thu, 21 Mar 2002 13:02:03 GMT

IS there anything specific to lookout for in the fcmsutil stat output to get a handle on whether the connection/card is having problems? The fcmsutil has been around for a few years now, but still its status output seems unintelligible and the lack of clear manuals/guide doesnt help either.
I normally grep for "-i error", but how to make out what value is good/bad? I have seen systems which have high error related values, but function fine.

Any clues on this one?

thanks
-raj

Re: fcms Driver state

Holger Knoppik — Thu, 21 Mar 2002 13:13:41 GMT

Hi raj,

did you try a fcmsutil /dev/RGDS, Holger

Re: fcms Driver state

Roger Baptiste — Thu, 21 Mar 2002 13:36:45 GMT

Thats exactly what i am looking for. to make sense out of those values. What is correct and what is not.

-raj

Re: fcms Driver state

Carlos Fernandez Riera — Thu, 21 Mar 2002 13:38:43 GMT

Loss of signal Count 0 Bad Tx Char Count 10

fm_ints 949

Bad CRC Count 0 Protocol Error Count 0

See for these parameters.
Some times a absolute value is less significant than a raisin value in sort periods.

Anyway bad CRC and fm_ints means an not absolute correct transfer, or in other words, bad signal.

Re: fcms Driver state

Roger Baptiste — Thu, 21 Mar 2002 17:14:21 GMT

fm_ints is around 800 for all the four fibre cards.
Bad Rx count is 10
Loss of signal count and Bad CRC count is 0

The system was rebooted today, so this are fresh statistics.

-raj

Re: fcms Driver state

Deshpande Prashant — Thu, 21 Mar 2002 19:12:57 GMT

HI
Did you check the kernel parameter
num_tachyon_adapters.

Normally I look for increasing "Bad Rx Char" value in less time frame in fcmsutil command.

Goodluck..
Prashant.

Re: fcms Driver state

Roger Baptiste — Thu, 21 Mar 2002 20:25:27 GMT

Hi Prashant! ;-)

Yeah, currently the num_tachyon is set to 5 and i need to boost it to 20. But, i am not sure whether it is connected with this problem, since it is a new occurence.

thanks
-raj