HP-UX 11i Oracle 8.1.7 on K260 chrashes

[Ingo Duschl]

Database stops/crashes regulary 24h.hello we got a DB running since 18 monthswithout failure.since 5 days the DB crashed about every 24h.archive log is on seems working.where are the logs to look on for UX and Oracle which are important, we can`t find the fault.we do a restart for the hole machine+oracle and it seems to run again about 24 hrs.please give advice where to start the search in the logs. THANX i guess a hardware thing not quite sure

[T G Manikandan]

To start with,

check the OS log /var/adm/syslog/syslog.log

check the db oracle log file -

alert.log

[Peter Godron]

Guten Tag Ingo,
First have to identify if OS or Oracle problem.
If oracle crashes, but machine does not show anything from dmesg or in /var/adm/syslog, I would check for oracle resource problems.
Has anything changed 5 days ago, more data on the system, additional processes?
Viel Glueck

[Keith Bryson]

Is this server low on memory (I have a suspicion)? Can you still ping the server, but you are unable to access the console or the server via telnet etc?

Keith

[Steven E. Protter]

This could be as simple as a full filesystem.

bdf

Is anything near full? I'd look at archive logs areas as defined in oracle.

It is possible hardware had a hand in it, like one disk of a disk array failing.

SEP

[Ingo Duschl]

we have changed the archive log destintaion because of full disks a week ago, but this seems to be working,made a new archlog-file after reboot today.but could be a reason. now there is enough space everywhere and everything was rebooted.(once a day ;-) )this was the only thing changed since month of good performance.
what i saw is: there are two disks in the arrays out of mirror,could also be the reason?!so got to bring back the mirrors first.and then we `ll see....
Thank you so far for quick reponse

[Jean-Luc Oudart]

Hi,

any error message in the alert log file ?
any dump file in the bdump or udump directory ?

Regards
Jean-Luc

[Indira Aramandla]

Hi Ingo Duschl,

Your oracle DB was functioning fine for the past 18 months without failure. You had file system full where the archive logs reside and so you changed the archive log destintaion.

And now your DB crashed every day. The first thing to check for Oracle database errors will be the alert logs and user dump logs located in the bdump and udump destination directories.

As Steven mentioned, it could be a simple reason where the archive log destination file system is full and so the database has suspended its activities. The database will resume its activities when there is some space for it to write the archive logs. Normally when you have the database in archive log mode, you need to have scripts scheduled through cron to compress the old archive logs, backup them up or copy them to a different location and clean up the old logs. This will help you to prevent file system full in that directory.

Can you check the alert log and see if there are any error messages and post them.


Indira A

[Ingo Duschl]

hello, so it seems to be a problem with a NFS filesystem, where old(one week and older) archivlogs are copied for backup and a oracle hidden parameter _controlfile_enqueue_timeout which was 900 ist set now to 2700. the errror message was:TIMEOUT ON CONTROL FILE ENQUEUE.It Was not of a HOT SPOT disk, the controlfiles are on 3 different disks which are not working much.The NFS filesystem is dropped so we will see if it is running........
thanks again for quick responses