cancel
Showing results for 
Search instead for 
Did you mean: 

Brrestore issue in HP IVM server

Former Member
0 Kudos

Hi

We are having an issue while restoring data from the backup. (SAP ECC 6.0 - Unicode / Oracle 10g / HP-UX 11.23)

when try to restore oracle datafiles from the backup of our production, all the files which are restored are 8192 Bytes less in size compared to our production backup, so when we startup oracle we are facing issues.

We have compared the filesystem properties itu2019s the same as production and also its set to largefiles=yes.

We tried restored the previous day's backups but all files restored are one block (8192 Bytes) less in size.

Say for example if the system datafile original size is 2147491840 Bytes the size after restoration in the target system is 2147483648 Bytes, which is 8192 Bytes less.

Our Target system is running in IVM (virtual host).

When we restore the same backup to a physical standalone system, the data is restored correctly.

Do we need to make any special changes at unix / filesytem level to make the restore files correct with exact size?

Given below is the fstyp of our IVM sytem filesystem

  1. fstyp -v /dev/vg01/lvol2

vxfs

version: 5

f_bsize: 8192

f_frsize: 1024

f_blocks: 125829120

f_bfree: 36536857

f_bavail: 34253304

f_files: 9134500

f_ffree: 9134212

f_favail: 9134212

f_fsid: 1073807362

f_basetype: vxfs

f_namemax: 254

f_magic: a501fcf5

f_featurebits: 0

f_flag: 16

f_fsindex: 9

f_size: 125829120

The init<SID>.sap file contents are

backup_mode = all

restore_mode = all

backup_type = offline_force

backup_dev_type = tape

backup_root_dir = /oracle/TU3/sapbackup

stage_root_dir = /oracle/TU3/sapbackup

compress = hardware

compress_cmd = "compress -c $ > $"

uncompress_cmd = "uncompress -c $ > $"

compress_dir = /oracle/TU3/sapreorg

archive_function = save

archive_copy_dir = /oracle/TU3/sapbackup

archive_stage_dir = /oracle/TU3/sapbackup

tape_copy_cmd = dd

disk_copy_cmd = copy

stage_copy_cmd = rcp

pipe_copy_cmd = rsh

dd_flags = "obs=64k bs=128k"

dd_in_flags = "ibs=64k bs=128k"

saveset_members = 1

copy_out_cmd = "dd ibs=8k obs=64k of=$"

copy_in_cmd = "dd ibs=64k obs=8k if=$"

rewind = "mt -t $ rew"

rewind_offline = "mt -t $ offl"

tape_pos_cmd = "mt -t $ fsf $"

tape_size = 180000M

tape_size_arch = 180000M

exec_parallel = 0

tape_address = /dev/rmt/1mn

tape_address_arch = /dev/rmt/1mn

tape_address_rew = /dev/rmt/1mb

tape_address_rew_arch = /dev/rmt/1m

expir_period = 2

tape_use_count = 100

volume_archive=(TU3A01, TU3A02, TU3A03, TU3A04, TU3A05,

TU3A06, TU3A07, TU3A08, TU3A09, TU3A10,

TU3A11, TU3A12, TU3A13, TU3A14, TU3A15,

TU3A16, TU3A17, TU3A18, TU3A19, TU3A20,

TU3A21, TU3A22, TU3A23, TU3A24, TU3A25,

TU3A26, TU3A27, TU3A28, TU3A29, TU3A30)

volume_backup=(TU3B01, TU3B02, TU3B03, TU3B04, TU3B05,

TU3B06, TU3B07, TU3B08, TU3B09, TU3B10,

TU3B11, TU3B12, TU3B13, TU3B14, TU3B15,

TU3B16, TU3B17, TU3B18, TU3B19, TU3B20,

TU3B21, TU3B22, TU3B23, TU3B24, TU3B25,

TU3B26, TU3B27, TU3B28, TU3B29, TU3B30)

Can anyone help on this regard

Thanks & Regards

Senthil

Accepted Solutions (1)

Accepted Solutions (1)

markus_doehr2
Active Contributor
0 Kudos

> Our Target system is running in IVM (virtual host).

That means you use HP Integrity Virtual Machine?

Note 1351051 - Support of Oracle for HP-UX Itanium Integrity VM

> When we restore the same backup to a physical standalone system, the data is restored correctly.

>

> Do we need to make any special changes at unix / filesytem level to make the restore files correct with exact size?

This is really a strange (and serious) problem. I could imagine, that the "virtual" disks created need a certain alignment which is, for whatever reason, not accomplished and so hence "cutting" the last block.

If you run the version mentioned in the note I would create an OSS call @ SAP for Oracle (BC-DB-ORA) and also involve HP into the process.

Markus

Answers (3)

Answers (3)

0 Kudos

HI

We have faced this problem recently. We highlighted this issue to HP,Bangalore and SAP. Till no reply.

After LTO 4 tape drive installed , we got this error. Earlier we have no problem with LTO 3. Server Details

HP-IVM- Production test server - ECC 6.0 - Oracle 10.2.0.4

recover database using backup controlfile until cancel;

ORA-00283: recovery session canceled due to errors

ORA-01110: data file 1: '/oracle/SFP/sapdata1/system_1/system.data1'

ORA-01122: database file 1 failed verification check

ORA-01110: data file 1: '/oracle/SFP/sapdata1/system_1/system.data1'

ORA-01200: actual file size of 44799 is smaller than correct size of 44800

blocks

Regards

Ravi.S

0 Kudos

Hi

Finally this problem was solved by HP Solutions Centre,Bangalore. Now we take the backup and restored in HP-IVM guest with dd_flags 64k. Now system working fine.

Regards

Ravi.S

markus_doehr2
Active Contributor
0 Kudos

> Finally this problem was solved by HP Solutions Centre,Bangalore. Now we take the backup and restored in HP-IVM guest with dd_flags 64k. Now system working fine.

Thank you for posting this. I hope SAP will add that to the IVM note.

Markus

0 Kudos

Thank you markus

I read sap note no.837670 . Any other note regarding this. Pl mention the sap note no

Regards

Ravi.S

markus_doehr2
Active Contributor
0 Kudos

I said, that I hope that SAP will add that information (about using a 64k blocksize) to the HP IVM note mentioned in my first answer.

Markus

Former Member
0 Kudos

Thank you Markus & Michael.

We reported the issue to SAP and HP. The issue is currently in their R&D division

We tried different permutations and combinations, finally when we take the backup from the single OS using the parameter

dd_flags = "obs=8k bs=8k"

dd_in_flags = "ibs=8k bs=8k"

in the init<SID>.sap file, then it can be restored in the IVM server with the same above parameter.

Even less than 8K works, but above 8K doesn;t work.

The disadvantage is it consumes lot more time to take the backup and restore.

Regards

Senthil

Former Member
0 Kudos

As mentioned above contact support.

Can you please post the output from both systems:

du -k <file>

ls -l <file>

cksum <file>

so when we startup oracle we are facing issues

What errors do you get?

What happens if you just copy a datafile to a virtual server? Are you using shared I/O(virtualized Luns) or attached I/O? On OS level please verify if you have the same filesystem (vxfs, jfs) versions and the same patch level.

Regards, Michael