Why is the db down ?

[chikkodi]

Is there a way to find out what caused the database to go down ? I looked into the alert.log file and there arent any errors reported there.

[tomate]

i don't think the db is down. Certainly you connect by the sqlplus on the server and on the server the ORACLE_SID is not set or set with an error.

Your question isn't very good. You don't say on which system it runs and OS etc.

Perhaps i go i the right way.

Regards
Thomas

[chikkodi]

Oracle version is 8.1.6 Enterprise Edition on sun solaris.

The database IS down because there are no background processes for the database, and I went into svrmgr and started it manually myself.

I posted this question because this has been happening for the last couple of days. One possibility is the unix servers are getting rebooted, but I want to be sure before I conclude thats the problem. Hence I looked into the alert.log file and there are no errors listed there. Is there any other place that can give me this information ?

[sambavan]

Check with your unix admin, the dates he/she had taken the system down. If that matches with your database down dates, then voilla, you have the answer.

If you are taking any cold backup on this instance, then check you crontab to see whether the backup script is broken, ie. it shutsdown the instance and would fail to bring it up.

Sam

[svk]

If the server reboot times are not conclusive enough, have your SAs look at system log to trace what is going wrong with the background processes.

[marist89]

If Oracle detects he needs to come down, he'll write something in the alert.log file (assuming the alert.log file is on a mounted file system). Check the uptime of the server to see when it was last rebooted:
% uptime

[sambavan]

Here is something that would be valuable to you all. I had these in my documentation, but couldn't/lazy to find earlier, inorder to include on my post, .

Hope this would help you:

Oracle always generates an audit file when the database gets shutdown. It can normally be found under $ORACLE_HOME/rdbms/audit. check the date/time stamp.

In the file, you may see something like this...

...
Tue Mar 6 11:50:30 2001
ACTION : 'connect internal' OSPRIV : DBA
CLIENT USER: oracle
CLIENT TERMINAL: pts/33

Tue Mar 6 11:51:50 2001
ACTION : 'shutdown'
...

If every user has a unique unix userid, then from there you can identify by the 'CLIENT USER' as shown above. But if you have a common userid to logon to your server, just like 'oracle' in our case, then refer to what's in the 'CLIENT TERMINAL', w/c is 'pts/33' in our case.

Using the 'client terminal' info, use the 'last' unix command to display from what host the user TELNETed from, as well as, the server logon time, as follows.

>last | grep pts/33
rsupport pts/33 dhcp-233113.ca.o Wed Mar 7 10:18 still logged in
rsupport pts/33 dhcp-233131.ca.o Tue Mar 6 12:29 - 14:26 (01:57)
rsupport pts/33 dhcp-233131.ca.o Tue Mar 6 11:45 - 12:13 (00:28)

Then 'ping' the hostname from your Microsoft Windows command line to determine the IP address of the origin host machine. Once you have the IP address, issue the command, 'nbtstat -A ' to determine the username and hostname the IP address is assigned to.


Sam