Hi managers,
after some extensive AdvFS and LSM-Operation (at least) one directory of
datafiles of our oracle-db 7.2.3 was unreadable and oracle refused to
start.
The directory-entry of what used to be a subdirectory looked like:
---------- 0 root system 8192 Jan 1 1970 btabi_6
Fortunately, we were able to restore from backup and the database seems to
be ok know.
Here is (in short) what I have done:
To enlarge the space of an existing AdvFS-domain I added a new LSM-volume
to this domain using addvol.
Next I removed all old volumes in this domain with rmvol causing AdvFS to
copy all the data (aprox. 30 GB) from the old volumes to the new larger
one.
After that I discovered the broken directory.
The server in question is a 2100 with DU3.2, all disk are rz29 rev. 0014
located in two external HSZ40 (HSOF V3.0). The disks are grouped to volumes
and mirrored by LSM.
My question:
Shouldn't I do this with AdvFS?
Is this a known problem with DU 3.2 (I recall reading a summery about AdvFS
being buggy in pre 3.2f-Versions)?
We had done this before on two other systems with no problems at all.
Is it possible that the file was already broken and the running oracle
database didn't noticed that?
Any opinions or experiences are welcome
Thanks,
Peter
mailto: Peter.Braack_at_degussa.de
Received on Wed Mar 19 1997 - 11:51:25 NZST