ORA-01110, 01115, 27072 during database startup

[BLADE_1]

hi,

Oracle ver. 8.0.5
OS - HP UX 11.0

I did a database cloning from my development server to another server..Both the platforms are identical.

The cloning was successful. Then suddenly after 3-4 hrs the database shutdown abnormally..I checked the alertlog and found the above ORA- errors. All these errors were pointing to datafiles in filesystems mounted on the same volume group VG03.

Last week we had the same problem so we thought that the disk is the problem and replaced the disk and did the cloning process.

We don't have LTO on destination server thats why we did NFS and copied the offline files using CPIO.

I thought it might be because of largefiles disabled but then the files which we were corrupt were of varying sizes 150 mb to 2 GB. And there are other files which are of 2 GB on the same filesystem which don't have this problem.

Anybody has any clues?

rgds
nainesh

[T G Manikandan]

Error tells that you have some problem with the disk on that VG.

I/O problems.


check the /var/adm/syslog/syslog.log for further messages on the same.
REvert with those messages

Thanks

[Christian Gebhardt]

Hi

It seems to be Bug 774252 which is fixed in 8.0.6.0 and 8.1.5:

<774252>
: ORA-27072 scanning across a 2Gb offset in a file
To determine if this is the problem that you are hitting we need to look at the
block number being reported with the ORA-01115 error. If this is multiplied by
the DB_BLOCK_SIZE and we are on the 2GB boundary.

Chris

[BLADE_1]

hi,
Done that...no error messages..This disk resides in a Hash box..we faced the same problem last week so we replaced it with a new disk..Funny thing is only some files are getting corrupted..
The filesystems on these vg are mounted and i can access them and even create some files in it..


We are trying to figure out if it could be the disk cage that is causing a problem..

rgds
nainesh

[BLADE_1]

hi christian,

ORA-01110: data file 59: '/oracle/D02/sapdata8/protd_2/protd.data2'
ORA-01114: IO error writing block to file 59 (block # 1)
ORA-27072: skgfdisp: I/O error

Filesize is not the problem..

If the bug was the problem then it should corrupt my other 2GB files residing on same filesystem as well as other filesystems..

rgds
nainesh

[T G Manikandan]

What are the permissions of the datafiles that you are trying to move.

YOu should have permissions 660 on the datafiles.

check the attachment

[BLADE_1]

hi,

The permissions and ownership are okay..no problem with that..rw-rw---- for all files..

rgds
nainesh

[T G Manikandan]

After these messages you should have got a HPUX error number.What is that?

Please post your logfile


Thanks

[T G Manikandan]

The problem could be with

1.permissions of the datafile(which is okay with your env)

2.Free space on the file system
Can the file system accomodate that file into that(sufficient amount of free space)

Revert

[Christian Gebhardt]

Hello BLADE

My error: Bug 774252 is only reported for SOlaris

nevertheless:

Note:49946.1
Note:62427.1
Note:62407.1
BUG: 749600

said that certain 'core' functions in Oracle and it's tools do not support large files (>2GB) and shows your errors (especially in 8.0.5 but even in 8.1.5).
So this errors may not arise continuously and on all datafiles greater than 2 GB.

Hint:
- not to use files larger than 2 GB
- Upgrade to 9.x

Note this:
11.0 8.0.5 64-bit Desupported
11.0 8.0.5 Desupported

Chris

[BLADE_1]

hi,

I have attached the log file..

rgds
nainesh

[T G Manikandan]

Looks like you are trying to or (from) copy to a non-
existent device(file system)

[BLADE_1]

This was working fine for sometime...if the files didn't
exist oracle wouln't have started at all..And believe me this filesystem exists..with the exact structure and permissions..I have tested this file system for read/write operations and its working fine..There is ample space in this filesystem and all the filesystems on this volumegroup..

I guess i have to wait for our server support guys to come and take a look..

[T G Manikandan]

Hi,

I would like you to check the /var/adm/syslog/syslog.log file for any errors on the disk or file system corruptions.

Please revert on the same.

[BLADE_1]

hi Mani,

I have attached the syslog.log file ..Please take a look

rgds

[T G Manikandan]

you have some problem with one of the scsi devices on bus0.

If you are running a new system then you should check the scsi cables and the termination.

If you had the system running fine for sometime it should be mostly a problem with a bad drive.

Looks lik e

bus0==c0
ssid=0a=10 on the bus

Looks like c0t10d0 is creating problems.

Just do a pvdisplay on the device.If it is not okay then it should show "stale" in the output.
#pvdisplay -v /dev/dsk/c0t10d0

//It is time to launch a call to HP and not Oracle//


Thanks

[BLADE_1]

pvdisplay is not showing any stale PE's. I know now that this is not an oracle problem..its pure hardware related. The filesystem has got corrupt which is mounted on Volume group vg03 and the device is c0t8d0..Its the same what we replaced last week..i guess its bad luck that the new disk is creating the same problem..Anyways i have logged the call with hp and am expecting someone to come anytime now..

thanx all of you for good inputs..

rgds
nainesh

[Indira Aramandla]

Hi Blade,
For the error ORA-27052 try this
Set these two values in your init.ora file.

DISK_ASYNCH_IO = FALSE
DBWR_IO_SLAVES = 0

and restart the database for the chages to be effective, anf then check the alert log for error message if they still appear.

[Leon Allen]

I've recently had a 'similar' problem.

eg. /s04/acuchis/dat/seatf.xdf : no such device or address........

Tell me, as a test, can you change permissions and ownership of, say, '/oracle/D02/origlogB/log_g14m1.dbf' (or one of your problem files) to something else? Then change it back? Does this test give you a "No such device...." error?

If it does give an error - then you've got a logical file error, and will have to recreate or restore from backup. The gottcha is, in anycase, I think your going to have to drop the logical volume and recreate it to get around this one (that's what I had to end up doing :-( )