vmunix: SCSI: Resetting SCSI -- lbolt: 7368431, bus: 5

[Arjun Kandasamy]

Hello


We have 2 x Vclass servers (Running HPUX 11.00 and Oracle 8.0.5 & SAP 4.0 DB - running dbs & executables) coupled with 4 x Nclass servers (Running HPUX 11.00 and Oracle 8.0.5 & SAP 4.0 DB - running executables only)

We have installed extra memory & disks (for swap). Intermittenty, intially and then consistently we are observing errors such as:

vmunix: SCSI: Resetting SCSI -- lbolt: 7368431, bus: 5
vmunix: SCSI: Resetting SCSI -- lbolt: 7368431, bus: 6

In addition on 1 of the Vclass servers we are getting errors such as:

vmunix: LVM: vg[1]: pvnum=4 (dev_t=0x1f05a000) is POWERFAILED

and

vmunix: LVM: Recovered Path (device 0x1f058000) to PV 0 in VG1


HP have advised raising the firmware level on all the new disks and some patch upgrades on all the systems as well as changing the disk controllers on all systems. This has been done and after 12 clean hours when system load was LOW, we have seen the above errors again on one of the N classes.


Any ideas?


AK

[Ross Zubritski]

AK,

Are the "power failed" volumes on a raid box? Is it by chance a Symm running power path?

Regards.

RZ

[Arjun Kandasamy]

No, local disks, mirrored....

AK

[Ross Zubritski]

Sounds like a disk going bad my friend. Powerfailed in my experience is not a good thing.

Regards.

RZ

[Arjun Kandasamy]

RZ

Yeah, that's what we are thinking here....an unhappy coincidence....

kind regards

AK

[Helen French]

Number of reasons for this message - SCSI, Patches, Firmware, H/W etc. Sometimes it cane be a simple LVM confguration too. You could just try increasing the "time out" value to start with:

# pvchange -t 180 /dev/dsk/cxtydz

[Arjun Kandasamy]

Tried this on one of the Nclass server increased to 120 and still no joy....

AK

[Helen French]

What about patches ? Do you have all new patches installed on the system? Are you using MC/SG?

[Arjun Kandasamy]

Not on the server that has failed if you mean Service Guard....HP engineer has turned up so we will see.....

Thanks...

AK

[Eugeny Brychkov]

These two disks are having lowest priority on the SCSI bus: Ids are 10 and 8. If system will swap hardly then these disks will not be accessible for this time. Increasing timeout as Shiju advised is a workaround, but may not work if bus is overloaded.
The solution in you case should be remove swap disks off this SCSI bus (generally: decrease 'continuous' bus load) or simply install one more SCSI controller and split disks between them.
Looking to the future you can think about fibre channel solution
Eugeny

[Steven E. Protter]

You have a disk going bad.

Though it could be a bad drive cage.

A bad scsi cable

A problem with the scsi card.

Or a bad unused drive elsewhere on the scsi chain.

All of these problems have bitten me on an old D-Class server over the years.

For certain you need hardware to come out and replace the missing disk.

If it contains important system info, do a make_tape_recovery first.

If by chance its an internal, hot swappable disk and you switched it out yourself, a reboot will make the lbolt go away.

SEP

[Ted MacDonald_1]

Some times this is an indication of not a bad disk but a bad power supply. Have the tecnician put a meter on the power supply. Some times it can be adjusted without being replaced. If it is the power supply your data will still be available once it is tweaked or replaced.

[CCIL]

The SCSI lbolt error are common. This errors generally comes if you have removed the external SCSI devices on line i.e when they are in power on condition , it could be external SCSI disk or Tapes .

For the LVM message for PV 0 can be ignored . If you are getting these messages quite often in the syslog , then the change the I/O timeout for each physical volume from default to 120 or to 180.

pvchange -t 120 /dev/dsk/cxtxdx

[T G Manikandan]

You should check the following

1.Make sure that the SCSI cabling and termination is OK.

2.Try increasing the timeout using pvchange.

If still the problem is there,then it is time to replace the disk.

[T G Manikandan]

Also make sure that your LVM patches are upto date.