Ken writes:
> At 11:20 AM 2/12/98 -0700, James Reynolds wrote:
> >This showed up in FSCHECK. What is it??? and is it catching??? and Is it
> >serious..
> >
> >Validating extent information.
> >*OVERLAPPING EXTENTS* DESCRIPTOR ADDR: $000001DC.$00263010,
> $000001D4.$002BA010
> >*OVERLAPPING EXTENTS* DESCRIPTOR ADDR: $00000011.$002DE610,
> $000001D5.$002B3110
...
>
> I talked to the RC about this and they were wondering if the FSCHECK was
> run while people were on the system. If is was, they suggest that you do
> it again with no one on the system and if you get the same messages to call
> the HP and have your disc drives checked for possible problems.
>
> FSCHECK is a pretty scary program to run with people on the system. I have
> only used the DISPLAYEXTENTS command with people logged on.
I agree...I try to avoid running FSCHECK while anything is happening on the
system.
The overlapping extents could, indeed, have been caused by file system
activity during the FSCHECK run.
It's possible, however, to determine for sure. One could take the extent
descriptor addresses shown and (via Debug/iX) backtrack to the file labels
for each pair of conflicts. Once you get to the parent file label,
the allocation date/time, creation date/time, and modification date/time
are available. That would provide strong clues as to whether or not
the files really overlap. (I.e., if the backtrack shows an unallocated
entry, then that file is gone now. if the backtrack shows a file created
about the time the problem was noticed, then the "active file system"
explanation is probably correct.)
Or...wait until the system is completely idle and run FSCHECK again.
If an idle-system FSCHECK shows the same overlapping extents, then
store off each of the associated files.
<plug>
De-Frag/X, from Lund Performance Solutions, can aid in the diagnosis
(and fixing) of overlapping extent problems.
De-Frag/X allows you to easily map from the reported overlapping
extent back to a file name, as shown in the example below:
De-Frag/X> displayextent $b9.$6db51c addr
(converted from extent pointer to file label @ $b9.$6db300)
-------------------------------------------------
Ldev 0, entry# 2 --> $b9.$6db300 = DEFRAGX.SOURCE.SIELER
File: DEFRAGX.SOURCE.SIELER (2 extents, label @ $b9.$6db300)
Ldev Disk Page# # Pages File Page # Extent entry@
---- ---------- ------- ----------- -------------
13 168,807 20 0 $b9.$6db510
13 266,748 2 20 $b9.$6db51c
# extents in file: 2
# pages in file: 22 pages
% fragmented: 4.7
Additionally, the DISPLAYEXTENTS (aka "DE") command always checks to
see if all of the pages of each extent are properly marked as
"permanent" pages in the disk's secondary storage bitmap.
http://www.lund.com
</plug>
Stan Sieler
[log in to unmask]
|
