checking if database is open on Unix

[leonard905]

Hi,

Is there a way to check if a database is open and available on unix without connecting to the database. I know that in 7.3 you can check for the sgadef*.dbf file and 8i you can check for PMON. But I also foud out that sometimes, if the db is mouted in restricted mode or was shutdown improperly a "ps -ef | grep ora_pmon" command will still show PMON even though users cannot access it. Even the sgadef*.dbf file sometimes is not deleted when the instance shuts down.

What's the most accurate method of performing the db check?

Thanks in advance,
Leo

[abhaysk]

Use DB Event Triggers to create a dummy file indicating its open.

Abhay.

[pando]

very easy, create some dummy/inexistent user and connect to database and look for ORA-01017 error like

a=`(sqlplus ksaks/kajsa|grep ORA-01017) << EOF
EOF
`

if [ "${a}" != "" ]; then
   echo Database is up"
fi

[abhaysk]

Originally posted by pando
very easy, create some dummy/inexistent user and connect to database and look for ORA-01017 error

Pando I think he meant even without attempting to connect to DB.

Originally posted by leonard905
Is there a way to check if a database is open and available on unix without connecting to the database

[pando]

without connecting you only bet is look pmon

[abhaysk]

Originally posted by pando
without connecting you only bet is look pmon

Why not with a trigger "After Startup ON Database" call some PROC where in you can output to a file some dumb info like (DB open with timestamp)& delete or rather append/flush the file with yet another dumb info like (DB Shutdown with timestamp) "Before SHUTDOWN"

Abhay.

[davey23uk]

what happens if your instance crashes, you will still have this file but no database

[abhaysk]

Ahh !! one of those rare case senarioz & ofcourse shutdown abort.


Abhay.

[marek_wiechula]

Checking the PMON does not help as far as telling whether the database is in restricted mode, a long running "shutdown in progress", an archival halt or an audit space freeze.

I think that any check for the status of the database will have to be more complicated than a single test.

For example, the check on the invalid user may need to be embedded in a background job so that the test can time out in a predictable manner if the database is in an archival hang. The invalid user will not work as far as detecting restricted session mode either because the password is checked before the restricted session privilege.

Perhaps a variation on the file idea but using DBMS_JOB to schedule an activity every minute (or 10 seconds or 5 seconds depending on how up to date you want to be) that uses FILE_UTL to update a file with a timestamp and the LOGINS field from V$INSTANCE... I am not sure if they would catch everything.

[TimHall]

Hi.

Whatever method you try it won't be as accurate as actually connecting to the database, so why not just do it. Apart from all the issues listed above there are sometimes issues like the archiver hanging which mean that the DB us up and running but you can't use it etc. All the tests you try will not guarantee that a user trying to connect will be able to unless you actually test using a connection yourself.

What is the reason for not using a test connection?

Cheers