Oracle froze

[Vogra]

Hi All!
I have Oracle 8.0.5 running on D270 server with UX-10.20. Today i had a different situation like this: the users was unable to connect and I only make connection with sys user (or internal). The oracle was freeze. The rollback segments was ok - tablespaces was ok. my only chose was a shutdown DB. After that, overall is ok.
In the alert file, I see:
--------
Fri Jun 22 12:53:31 2001
>>> WAITED TOO LONG FOR A ROW CACHE ENQUEUE LOCK! pid=20
--------
The message repeated some times.
I think it's efect, not cause of the problem.

Some traces files was generated.
After my initial investigate, I not discover anything.
I have some development ocurring im my production system (I know it's not recommended, but...).
Some ideas?
I can send more information...

[Eileen Millen]

Are there any messages from around that time in
the /var/adm/syslog/syslog.log file

Eileen

[Rita C Workman]

Did you get an oracle error number to reference??

Rita

[Rita C Workman]

Here is some additional info I found regarding your enqueue error:

The session is trying to get a data dictionary lock.

Wait Time: Wait up to 60 seconds.

Parameters:

cache id
The CACHE# column value in the V$ROWCACHE view.
mode
See "mode" on page A-8.
request
The pipe timer set by the user.

Here's the url where I found this ( use the little index icon at the top to look for more):
http://www.csee.umbc.edu/help/oracle8/server803/A54645_01/events2.htm#211464

Hope this helps,
Rita

[Vogra]

Hi!
in the alert file, after shutdown, the following lines was written:
Errors in file (...)/ora_7733.trc:
ORA-07445: exception encountered: core dump [10] [2063849312] [1073755912] [292] [] []
ORA-01089: immediate shutdown in progress - no operations are permitted

and the BD was closed normal.

[Rita C Workman]

Check your init.ora for the following response on the one oracle issue ora_01082, I found this suggested action to take for the following:

ORA-01082: ROW_LOCKING = ALWAYS requires the transaction processing option

Cause: ROW_LOCKING = ALWAYS is specified in INIT.ORA file. This feature is not supported by ORACLE
without the transaction processing option.

Action: Remove ROW_LOCKING = ALWAYS from the INIT.ORA file or set it to DEFAULT or INTENT.

Hope this helps,
Rita

[AlHassani Mohamed]

Hi
This is an oracle Problem that only occurs on
oracle 64 Bit. But HP-UX d?snt support 64 Bit.
You mus?t verify which Oracle you have installed. If the is like about descrobed, you must deinstalling that and reinstall oracle 8.0.5 32 Bit for HP-UX 10.20.
regards
Mohamed

[Vogra]

Millen,
nothing in syslog...

Hi Rita,
I don't have this parameter in init file...

Mohamed, I'm running Oracle 8.0.5 32 bits on d270 (hp-ux.10.20)...

Hi ALL,

I'ts possible to monitor locks, ok? How can I do this? I heard something about lock manager. What about ?
Thanx in advance.

[Steve Slade · Accepted solution]

Claudio,

Lock Manager is usually bundled in with Oracle enterprise Manager. Its quite a nice GUI tool. If you do not have access to this, you can try the Oracle Utility script utllockt.sql - which is in $ORACLE_HOME/rdbms/admin. It has been a while since I have used it, but I seem to remember that it was ok - it will give you a tree view of the locks on the system.

However unless the problem is re-occuring, it will probably not be very useful. I would have a look at the trace files. Most of it will be meaningless - or at least seem that way. However, usually the first page or so can give you an idea what was running at the time. As the lock seems to have been on the system tables. Apparently, a number of Oracle operations on the system tables are serial - such as extent allocation. Could it be that something was dynamically adding extents,(maybe a number of small extents) to a table and locked everything else out?

Other than that - the only time I have ever seen Oracle Freeze is when running in archive log mode, and the archive log destination directory is full. The Archive process stops, the available logs fill up and everything grounds to a halt.

Good luck!

[Vogra]

Steve,
after a long time, I'm be back. Well, I'll use the script did you hint to me.
Although I'm not right about cause of this problem, I'm inclined about a developer with a specific application and I had been monitoring (trace, etc)him.
Thanx so much!