disk power failed

[Javier Ortiz Guajardo]

In document KBRC00000300 explains the causes of POWER FAILED as follow:
LVM may have had problems accessing the disk because of a high I/O load or possibly the disk is not properly terminated. It is also possible that the disk is truly going bad.

i do not understand what means the disk is not properly terminated. and how can i resolve or detect if my disk is in this cause.
con somebody explains this, please.?
thanks

[Jean-Louis Phelix · Accepted solution]

Hello,

I think they are talking about SCSI terminators. One missing could lead to I/O errors.

Regards,

Jean-Louis.

[jim bidebo]

have a look at this pretty complete guide on scsi and termination : http://www.seagate.com/support/kb/disc/termination.html

[Ashwani Kashyap]

This could also be as a result of I/O's to the disk timing out ( when there si heavy I/O ) .
If this is the cause then this can be solved by increasing the disk timeout value . Specailly in disk arrays like emc etc . its good to increase the disk timeout from default to 180 sec .
#pvchange -t 180 /dev/dsk/c.t.d.

Try doing this and see if the message still comes . If it still persists , there could be other hardware problems .

[A. Clay Stephenson]

Regardless od Single-ended, differential, FWD, LVD, ... SCSI a few rules must be obeyed.

Termination refers to a network that looks something like this for each data line on a SCSI bus:

+5 Volts --- resistor --- data line - resistor --- Gnd

The datails vary a bit between the various SCSI versions but the idea is that a signal line is 'suspended' between a +V and -V rails. If resistors are used this is called 'passive' termination and if transistors are used to emulate the resistor network, thisis called 'active' termination.

Now the whole purpose of this is to reduce 'ringing' on the bus. The idea is that a pulse hits this termination network and is not reflected much like a ball hitting wall of cookie dough.

Two conditions have to be met. 1) There must be a terminator on each end of the bus and ONLY at the ends. Typically, HBA's terminate on the card itself. 2) There must be termination power supllied to the bus by at least one device on the bus. This provides the +V and -V that the termination network requires. The power can come from any device.

You need to check that you have exactly two tewrminators and that they are on the ends of the bus. One of the best ways is to replace one of your terminators with a terminator that displays bus status. It will indicate that the bus is properly terminated and powered.

Having said all this, if your drive has been working well and is not failing occasionally, it's quite probable that your drive is actually failing. Nonetheless, check all cables to make sure that they are well-seated.

[S.K. Chan]

The POWERFAILED error can mean a number of things depending on what other error messages comes with it.
For example if you see something like ..
"pv[#] returned to vg[#]" this is more likely due to low timeout value (which you can change by pvchange -t). Another example .. say if you see something like .."SCSI : Request Tmeout--lbolt--" then this is more likely due to SCSI termination, bad SCSI card or disk going bad. When you see this error, it'll usually indicate the device number , like so (example)..
0x1f052200 which translate to c5t2d0
That way you would know what to check for. Disk that are not properly terminated could mean you did not terminate you SCSI chain at the end or the SCSI terminator is loose, ie not properly connected. Regardless of what kind of PWERFAILED error messages you get, you would still want to try eliminate all possibility like ..
- Run diagnostics to make sure it's not the disk.
- Change io timeout value if you suspect it's due to low timeout.
- Check cable/connection/termination.

[Eugeny Brychkov]

Ask Arnold Schwarzenegger to finish the disk ;-))

[doug hosking]

In addition to several other good ideas posted,
I seem to recall that this error can also be returned in some cases if your drive does not have the correct firmware. Exactly which version you should have depends on the particular model of disk drive you are using.
You might want to run 'diskinfo -v' on the disks in question, determine the model number and rev level for each, and ask your HP rep if
the firmware on the drive is current for that
model. For example:

$ cd /dev/rdsk
$ diskinfo -v c2t13d0 | more
SCSI describe of c2t13d0:
vendor: HP 18.2G
product id: ST318452LC
type: direct access
size: 17783240 Kbytes
bytes per sector: 512
rev level: HP02
blocks per disk: 35566480

shows a Seagate ST318452LC drive with
firmware revision HP02.

With disk models changing so fast, I won't even pretend to have all the relevant information for each model you might have, but hopefully your HP support rep can look it up for you.