cancel
Showing results for 
Search instead for 
Did you mean: 

ORA-17612 during DB instance install

Former Member
0 Kudos

hi oracle guru's - i have hit this error when doing doing a db instance install:

ORA-17612: Failed to discover Oracle Disk Manager library, return value string

Cause: Discovery of the odm library by calling odm_discover() failed

Action: Contact your storage vendor who has provided the ODM library or call Oracle Support

i am using sapinst to create the DB instance, but can reproduce the problem by running the same commands sapinst is doing behind the scenes (i.e. running "startup nomount" in sqlplus throws the error)

some background info is relevant, as ODM (Oracle Disk Manager) and VxFS filesystem setup is involved. i have that and some screenshots below - please let me know if you have any idea how to get past this - thanks

==

we are bw3.5 on hpux11.23(ia64) and ora10202, sid is u201CBW4u201D. this is our UAT system and we are in the

middle of the DT window for a Unicode conversion

we have a ServiceGuard cluster w/ 2 nodes:

g1u0340c

g1u0341c

the DBCI package runs on either node. currently the db is running on

340c node. so the 340c node is the non-unicode (NUC) node. and the

way we are doing this UC conv project is to essentially "break" the

HA/SG cluster temporarily (no failover capability during the tech

conversion)...and export from the 340c (nuc) node into the 341c node

(unicode node).

since this was a SG setup, we had the oracle binaries/patches/everything installed on both nodes already. (i.e. all

the normal installation dir's under /oracle/BW4/102_64 existed - on

both nodes - and still do).

one other point that i must bring up, due to the specific error #

(ora17612) is ODM. we have ODM (asynch io) setup here on BW4 on both nodes of the cluster. the ODM was installed a few months ago but (in hindsight) we failed to do proper failover testing of the SG cluster to see if oracle would run on the 341c node.

the ODM setup/installation on BW4 happened on both nodes of the cluster

and involved these main items (much from note 999524):

- removing T2795BA (v. A.01.00 of HP Serviceguard Cluster File System

with MCOE), then installing T2796BA (HP Serviceguard Cluster File

System for Oracle) [by Unix team]

- adjusting the dbci<SID>.control.script, a SG config file (making some

mounts have 'noqio') [by Unix team]

- other mounts were chg'd to noqio based on this: if CFS (then chg'd

at command line)...if not, then these dedicated mounts were changed

in /etc/fstab [by Unix team]

- config soflink from oracle files to ODM library (in /opt/VRTS*)

in /oracle/<SID>/102_64/lib [by DBA per note 999524]

so now you have all the "background" info...now on to the actual error

i got and where i was in the SAPINST DB INSTANCE install when i

received the error. 1 more bit of "background" - as far as this DB

INSTANCE install...my goal here was to "reuse" the existing oracle

installation and ~46 one-off patches (since that is not specific to

unicode/non-unicode i should be able to do that). so...that means when

sapinst got to the prompt telling me to go install the oracle software

(via OUI)...at that point i didn't have to do anything (again, all the

normall installation files/dir were already there on the 341c node

under /oracle/BW4/102_64)...So at that pt i just hit 'ok' at the

sapinst prompt and it went on to the next steps (primarily, to create

the new database)

sapinst threw the ora17162 error during the u201Ccreate oracle databaseu201D stepu2026

so it was logging in via sqlplus and doing a u201Cstartup nomountu201D when the error occurred.

I have compared both nodes and cannot find any difference in the ODM related files that I know of at leastu2026that is stuff in /opt/VRTSodm as well as stuff in the /oracle/BW4/102_64/lib (the softlinks from the library files there to the /opt/VRTSodm directory)u2026I also cannot see any difference in the requ2019d HP SG SMS software that was required (swlist matches for the expected HP SG SMS software)u2026so Iu2019m at a loss to figure out where oracle is looking and canu2019t find the ODM library u2013 if you have any ideas what I could check please let me know!

(apparently oracle runs a "odm_discover" call b4 during startup...and this is where it's failing)

Accepted Solutions (0)

Answers (1)

Answers (1)

Former Member
0 Kudos

Hello Ben,

I don't know if it will be usefull but have you checked this note from HP yet?

http://docs.hp.com/en/T1859-90038/ch02s17.html

Maybe it will help you to find what's going on.

Former Member
0 Kudos

hi sebastian - several hrs of digging into it and we found the problem - it was more of a problem we had bc we had CFS filesystems related to odm that weren't mounted - once that was fixed the error no longer occurred

Former Member
0 Kudos

Nice to hear that you found the solution!

Regards,

Seb