Re: ORA-27101 hell!!!

Chris Clarke wrote:

> Hi all,
>
> I've installed Oracle 8.1.7 EE on RedHat 7.2 (as per install guide,
> v2.1 gcc libs etc.)
>
> I managed to create an instance "O817DB" using dbassist.
>
> I started the instance as follows:
>
> $ sqlplus internal
>
> SQL> startup
>
> The startup procedures gave no errors.
>
> I then tried to get into the DB as follows:
>
> $ sqlplus system/manager
>
> No problems there either. I could run queries, create users, etc.
>
> Now I tried to start the TNS listener:
>
> $ lsnrctl start
>
> Doing:
>
> $ lsnrctl status
>
> ...revealed that I had 1 service handler for O817DB as expected.

Difficult to say: the listener can be 'aware' of things which don't actually exist. If you statically declare the existence of your instance in listener.ora, then even if the instance isn't running, lsnrctl status will report that the listener is handling that service. It's status will be 'UNKNOWN'. If you have automatic instance registration working, then the listener will report a status of 'READY' only when the instance is actually running. And if you have both AIR and static declarations in the SID_LIST, then you should see 2 entries for the service, one with a status of UNKNOWN and one READY ...provided the instance is actually running.

Now 27101 usually means that the instance isn't running... yet you just connected to it without using TNS, and it seemed to be working fine.

So, what I would do is: when you get your 27101 error, immediately do a ps -ef | grep XXX, where XXX is your ORACLE_SID, and see whether you have listed the various background processes (lgwr, arch, dbw0 and so on) which indicate the healthy existence of an instance.

If you get no processes listed, then you have to ask why the instance just died after working perfectly for system/manager. Usually, it's because kernel parameters (shmmax and sem) haven't been set properly. If you do the 'echo 250 32000 100 128 > sem' stuff they tell you about in the installation doco, that's only valid for the lifetime of your box, and a machine reboot will reset the parameters back to inadequate Red Hat defaults. You need to edit sysctl.conf to make the changes permanent.

If you *do* get processes listed, then it indicates a problem with your TNS setup. That might be listener mis-configuration, or it might be tnsnames.ora mistakes. We'd need to see both to determine that.

You might find my paper on installing onto Red Hat 9 of use (i know it's the wrong Oracle version and the wrong Linux version, but the installation bit only takes 20 pages or so. The rest of the document walks through setting up networking, and that might be the issue here).

Regards
HJR

-- 
--------------------------------------------
See my brand new website, soon to be full of 
new articles: www.dizwell.com.
Nothing much there yet, but give it time!!
--------------------------------------------