Tru64 Rebooted suddenly

[SCSI Error]

Hi Could anyone interpret this output from CCAT?

Thanks


---------- - rbf_pin_record: negative offset: pg addr N1, rec addr N2\n N1 = 1618804736, N2 = 83886115 Digital UNIX V4.0F Node: hostname ----------

Full Description:


---- Number of Rules Matching this Case ----
Rule Match Count: 0

---- Source Rule Info. ----
Source Rule Set: Tru64_Unix_RULES_Generic: 8/18/2003
Tru64_Unix_RULES_V40F: 8/18/2003

---- Rule Match Results ---
Status: UNIDENTIFIED

Evidence:
Tru64_Unix_Main Tru64_Unix_Generic Tru64_Unix_V4.0F

PHYSICAL_MEMORY: 3071
STACK_TRACE: stop_secondary_cpu panic event_timeout xcpu_puts printf panic advfs_sad rbf_pin_record dealloc_bits_page dealloc_bits_no_sub_ftx sbm_return_space_no_sub_ftx del_range del_xtnt_array del_dealloc_stg close_one_int close_int bs_vfs_close msfs_inactive vrele msfs_remove unlink syscall _Xsyscall
RETURN_ADDR_I_MODULE:
CRASH_TIME: 11/03/2008 06:49:25
PANIC_STRING: rbf_pin_record: negative offset: pg addr N1, rec addr N2\n N1 = 1618804736, N2 = 83886115
PC_I_MODULE:
AVAILABLE_CPUS: 2
SAVED_EXCEPT_FRAME_PTR:
ARCHITECTURE: axp
UPTIME: 242.83 hours
HOSTNAME: host
CRASH_ANALYSIS: gui
KMF_FAULTING_PC: 0x0
PANIC_CPU: 1
FAULT_VIRT_ADDRESS: 0000000000000000
FIRMWARE_REV:
OS_VERSION: V4.0F
SYSTEM_STRING: ES40
NUMBER_OF_CPUS: 2
OPERATING_SYSTEM: Tru64 Unix

************ End of Message ************

[Venkatesh BL]

"rbf_pin_record: negative offset: pg addr N1, rec addr". Its an AdvFS error. Seem to have occurred while removing a file.

[Rob Leadbeater]

Hi,

I'll go along with that...

The references to "del_xtnt_array" and "del_dealloc_stg" would also seem quite relevant to your other post on BMT extents...

I suspect one of your file domains could be quite poorly. What sort of files are on the file system ?

Cheers,

Rob

[SCSI Error]

there are number of small files in it with just below 100 bytes each files.

i looked into cron. the time rebooted was the same time with the cron runs. The cron script tar the small files into 1 file and delete those files.

Does this cause the issue? why it panic'd?

[Martin Moore]

You may have metadata corruption in that domain. I suggest you run 'verify' against the domain. You can also try 'fixfdmn' if you're on patch kit 7 or later for 4.0F.

Martin

[Rob Leadbeater]

Hi,

> there are number of small files in it with
> just below 100 bytes each files.

How many is a "number" ?

Early versions of AdvFS didn't handle *lots* of small files in a domain particularly well...

In addition to the suggestions to verify and fixfdmn the domain, you may also want to think about migrating the data to another domain. Check out the man page for mkfdmn and read up the section on the -p and -x options.

Cheers,

Rob

[SCSI Error]

around 100K # of files just in one directory not including other directories. These where the small files resided.

I deleted files and brought back to 5K. I didnt see any changes on extentCnt after deletion of files. Will it not automatically decrease once I deleted files?

Thanks

[Rob Leadbeater]

No. You will have to recreate the domain.

Cheers,

Rob

[Martin Moore]

It might not be necessary to recreate the domain. If you have some spare disk space laying around, you could possibly use addvol with the -p and -x options to add a new volume with a larger BMT to the domain. This may help alleviate the situation, especially if you can then rmvol the original volume(s).

Martin

[SCSI Error]

Do fixfdmn reduce the value of extentCnt? coz I already delete the small files.