Oracle8

[Mark_596]

We have oracle8 running on our unix box. I dont know anything about oracle.

But today for the first time, the database decided to shutdown. Can anybody tell me if there are any log files on why this happened and more importantly who.

Thanks

[RAC_1]

oracle keeps alert.log. You may have a look at it. It is loacated somewhere under oracle directory.

Anil

[Steven E. Protter]

/oracle/product/8.1.7/rdbms/alert.log

If your version is 8.1.6 then change the 8.1.7 to 8.1.6

This is based on standards set by oracle.

if your instance name is datafile:

/oracle/product/8.1.7/rdbms/alertdatafile.log

Don't do any database operations as root. Do them as the user of the binaries.

SEP

[Chris Xu]

As root, switch to the owner of Oracle DB, normal 'oracle".

$cd $ORACLE_HOME/rdbms/log

Take a look at the file named alert_.log, and should find some info there.

[Mark_596]

I have just checked the alert.log file but nothing in there for the last couple of months.

Is there any more logs?

THanks

[Chris Xu · Accepted solution]

It sounded like the active alert.log was somewhere else. To comfirm that, look at the timestamp and it should be close to the time your db went down. Now older versions of Oracle 8 may have the alert.log under a directory like:

//app/oracle/admin//bdump

[Steven E. Protter]

Search the entire system for the log file. in the init.ora file there may be a statement redirecting the log.

If you have found the log and there is nothing in it, I'd suggest looking for core dumps.

Lastly, I'd check the filesystems oracle is sitting its data on. If they are full, this may be as simple as adding some filesystem space.

Something caused it to fail, but it depends on the issue whether or not you can just boot the server and move along.

SEP

[Mark_596]

Found the logs like you said they where somewhere else.

I have noticed in the logs that the following is mentioned

Alter Database backup controlfile to trace
Time stamp
Completed: alter database backup controlfile to trace

just before each night when the backup goes in.

The above is also mentioned when the database went down. Does that mean somebody has run our backup script and realised what they had done?

Can we tell what user has shutdown the db?

Thanks

[Patti Johnson]

Mark,

Alter database backup controlfile to trace;
is not a destructive statement - you can run this at anytime to create a text version of your controlfile contents - it is part of most standard backup scripts though.
If a normal cold backup procedure was run you should see something similar to
"ALTER DATABASE CLOSE NORMAL" in the log as well. This would tell you that the db was intentionally shut down.
If you found the alert_.log file in a directory named $ORACLE_BASE/admin//bdump you may also find some trace files in $ORACLE_BASE/admin//udump that could indicate if the db crashed.
The ORACLE_BASE environment variable is usually set to the login directory for the oracle account - but it could be anywhere.

Check your sulog to find out if another user did a switch to the oracle account - also check with any user who is part of the same group id as the oracle account ( usually DBA ) they would have permission to shutdown the database.

Patti

[Chris Xu]

You can't tell exactly which user has shutdown the db if the db did not go down due to abnormal reasons. Anybody who has access to the db owner account could do it. The best you can try is to look at the sulog, and use "last" command to find users who logged on from where around that time frame. And look also at users' .sh_history if that is enabled.

[Mark_596]

Again thanks

I will do some more investigating tomorrow. The reason why I asked about the backup. Was because I asked our DBA to check the backup script to ensure everything is in there what we needed.

follows more info from the log file.

Shutting down instance (normal)
License high water mark
time stamp
shutdown waiting for logins to complete
time stamp
shutting down instance (abort)
timestame
starting oracle instance (normal)

Thanks

[Steven E. Protter]

EEEK!

After any shutdown abort, you need to bring the database back up again and shut id down normally.

shutdown abort

.... messages

startup

.... messages

shutdown immediate

Also, you may have a licensing issue on your hands. WE still have Oracle 8 and I don't ever see that message.

SEP

[Patti Johnson]

Mark,

From what you posted from the alert log you might have a concern - if the backup is done after the shutdown (abort).


Shutting down instance (normal)
License high water mark
time stamp
shutdown waiting for logins to complete
time stamp
shutting down instance (abort)
timestame
starting oracle instance (normal)

If you do a shutdown - without any option - normal is the default. This means that Oracle will wait for all connected users to logout. If they leave a session open - oracle will wait and wait... regardless of activity.
The shutdown abort will bring the instance down quickly - but the database will not be is a consistent state to do a cold backup.

A better option is to do a shutdown immediate, oracle will then cancel all active session and rollback the transactions before shutting down. The database is then in a consistent state and your backup is valid. If you must do a shutdown abort ( because the shutdown immediate hangs ) then do a startup and shutdown immediate before running your backup.

Patti

[Yogeeraj_1]

hi,

i would check the logs at the OS level also. If there is no messages about a shutdown in the alert.log then it must have been some sort of crash at the OS level!

did the server restart?
Are there any core dumps?

i would check the shutdown.log and also verify if the database is configured to start automatically or manually...

just some thoughts.

can you post your alert.log?

just some thoughts

regards
Yogeeraj

[Jean-Luc Oudart]

Hi Mark,

if someone connect as SYSDBA (or internal) he/she will be logged into the audiot trace file.

The audit trace file is located in $ORACLE_HOME/rdbms/audit

cf . Metalink Note:103964.1

Regards
Jean-Luc