frecover errors when run via at

[Peter Heinemann]

Good Day,
Anyone know why the following command, which is copying a file from one file system to another (disk to disk)

fbackup -v -i /db/oracle00/rgdwp/ps_cert_apr01.dbf -n -f - | (cd /db/neworacle00; frecover -Xsrf - )

works fine interactively, but fails as a background job (via at) with the following error?

fbackup(1004): session begins on Fri Jul 11 10:11:48 2003
fbackup(3024): writing volume 1 to the output file -
1: / 2
2: /db 2
3: /db/oracle00 16
4: /db/oracle00/rgdwp 16
frecover(1047): I/O error in write of file /db/oracle00/rgdwp/ps_cert_apr01.dbf
5: /db/oracle00/rgdwp/ps_cert_apr01.dbf 4194368
frecover(1001): file trailer does not match file header for
frecover(1002): file /db/oracle00/rgdwp/ps_cert_apr01.dbf
fbackup(1005): run time: 63 seconds
fbackup(3055): total file blocks read for backup: 4194404
fbackup(3056): total blocks written to output file -: 4194434

Remember, it works fine interactively so there's apparently nothing wrong with the frecover syntax.

Thanks...

[Bryan D. Quinn]

Hello,

What is the exact syntax that you are using with the at command?

-Bryan

[Massimo Bianchi]

Hi,
here is the problem:
4194368


this is the maximum number of block readable by the ksh and by the at command, otherwise specified.

To have it work:
1- specify a shell different from ksh, like /sbin/sh or csh
2- change /var/adm/cron/.proto
the line
ulimit #######
to
ulimit unlimited

HTH,
Massimo

[Steve Steel]

Hi


Add the latest patch

hp-ux_patches/s700_800/11.X/PHCO_25424 :fbackup(1M):frecover(1M):cumulative:
Target: 11.00


Could be too many devices on the same scsi as the tape . This program should be interactive so maybe the lack of a real tty is killing something


Steve Steel

[Peter Heinemann]

Massimo:

I'm already using posix (sh), not Korn.

Brian:

it was:

at -f scriptfilename now

[Massimo Bianchi · Accepted solution]

Hi,
then check /var/adm/cron/.proto
both requirement must be fulfilled.

HTH,
Massimo

[Peter Heinemann]

Steve:

it's a disk-to-disk backup (see my original post), not tape, and the disks are SAN not scsi.

I'll check the patch anyway.

[Peter Heinemann]

Steve,

I neglected to mention that we're at 11.11, so the patch you suggested doesn't apply.

[Massimo Bianchi]

Hi,
since you are using largefiles (it looks like so), check if destination lvol can use largefiles:

mkfs -Fvxfs -m /dev/vg../lvol

and check if largefiles is there.

HTH,
Massimo

[Peter Heinemann]

Massimo - 10 points to you. Definitely the .proto

with ulimit $l -- copy failed
time(seconds) unlimited
file(blocks) 4194303
data(kbytes) 1048576
stack(kbytes) 131072
memory(kbytes) unlimited
coredump(blocks) 4194303
nofiles(descriptors) 1024

with ulimit unlimited -- copy successful
time(seconds) unlimited
file(blocks) unlimited
data(kbytes) 1048576
stack(kbytes) 131072
memory(kbytes) unlimited
coredump(blocks) 4194303
nofiles(descriptors) 1024

at within the interactive shell -- copy successful
# ulimit -a
time(seconds) unlimited
file(blocks) unlimited
data(kbytes) 1048576
stack(kbytes) 131072
memory(kbytes) unlimited
coredump(blocks) 0
nofiles(descriptors) 1024

So, what fills the $l parameter?

[Massimo Bianchi]

Hi,
$l
is passed from the at daemon to the script, and is evaluated as a default to that number of blocks.

You can see the vaule with a sample script that does an

env > /tmp/curiosity.txt

and you can see how mush stuff goes there !

I learned the existance of this file and config when my oracle had errors ora-600 when started from at....

HTH,
Massimo