[Search for users]
[Overall Top Noters]
[List of all Conferences]
[Download this site]
| Title: | VAX and Alpha VMS |
| Notice: | This is a new VMSnotes, please read note 2.1 |
| Moderator: | VAXAXP::BERNARDO |
|
| Created: | Wed Jan 22 1997 |
| Last Modified: | Fri Jun 06 1997 |
| Last Successful Update: | Fri Jun 06 1997 |
| Number of topics: | 703 |
| Total number of notes: | 3722 |
225.0. "Very unusual issue with backup (fyi)" by CSC32::R_GROSSMAN (Don't forget to dance.) Wed Feb 19 1997 17:42
Here's an interesting story (check me on the plausibility,
but, I believe this is what happened).
It all started with a customer reporting that backup was backing
up way too much data. When we looked closely we found a file in
the backup listing with an empty directory spec, e.g. []filename.
ANALYZE/DISK reported absolutely no problems with the file
structure. If we took off the /IGNORE=INTERLOCK we got a
"system-f-filacccon, file access conflict" (btw, it was an image
backup). Looks like the file is delete pending and still open
by some process. However, we shutdow the system and it's still
there. Normally, the customer drops out 1 member of a 3 member
shadow set for their image backups.
So, the theory is (1) the customer deleted this 3 million block
file and there were channels still open (2) they dropped a member
out of the shadow set and performed their image backup in this
state (e.g. delete pending with a nonzero reference count) (3) they
restored the disk and now we're corrupt. We have a file not in any
directory that is delete pending with a nonzero reference count,
but, no channels assigned.
The only thing that surprises me is that ANALYZE/DISK doesn't
report any problem whatsoever. I'm pretty confident that this
was the scenario. It's just very weird. Murphy's Law I suppose.
Just thought this might be interesting and a warning to
simply dropping a single member out of a shadow set for
backups.
Regards,
Russ
| T.R | Title | User | Personal
Name | Date | Lines |
|---|
| 225.1 | | AUSS::GARSON | DECcharity Program Office | Wed Feb 19 1997 18:30 | 31 |
| re .0
>When we looked closely we found a file in the backup listing with an empty
>directory spec, e.g. []filename.
It may be a temporary file (that is created without ever being in a
directory). Are they doing large SORTs?
>now we're corrupt. We have a file not in any directory that is delete pending
>with a nonzero reference count, but, no channels assigned.
A file on disk does not have a reference count (of channels on which it
is accessed). There is in memory information for which your statement
could be true but I doubt it.
In any case if you have shutdown the system then your statement is
unlikely to be true (since the in memory information is lost).
ANAL/DISK would report the file if it were delete pending (i.e. marked
for delete).
A file created with the RMS option 'TMP' is never in a directory and
nor is it marked for delete and so if the creator is interrupted in its
processing, the file will be left orphaned (not in a directory) but not
marked for delete.
A file created with the RMS option 'TMD' is never in a directory but is
marked for delete. If the system is interrupted (e.g. system crash) the
file will be left orphaned and marked for delete.
ANAL/DISK should report a file that is not in a directory.
|
| 225.2 | | VMSSG::FRIEDRICHS | Ask me about Young Eagles | Thu Feb 20 1997 10:31 | 14 |
| When they "drop the member out of the shadow set", are they
dismounting the whole set, then remounting just the remaining
members?
If not, they should be... A member that is dropped out of a
shadow set will not necessarily be updated with the latest operations
that are in the cache. There are numerous notes in here talking about
this.
Corruption possible?? You bet.
Cheers,
jeff
|