Database crashes

[SpiritOfLennon]

Hi,
I'm hoping someone can help me with this as I am rapidly ageing. We have an Oracle 8.1.7 database on a HP UX 11.00 server. The database seems okay most of the time and then suddenly, seemingly for no reason the load on the Unix server goes sky high and the database grinds to a halt. There are no intensive Oracle process or Unix processes working and HP have looked at the server and claim it is okay. If anyone could give me any advice on where to look and what the problem may be it would be highly appreciated. Just to give you an indication of my level of experience I have a fair knowledge of Oracle DBA skills and good HP -UX skills. Any suggestions including humourous ones would be appreciated at this stage as people are starting to scream ( it's month end ). SOL
I'm only guessing but my guess work generally works for me.

 

[trunix]

I have seen this happen before, but cant fully recall the reason. I seem to think it was something to do with shared memory although I am not 100% sure about this.

I think we used ipcs to see the problem and ipcrm to remove, but dont take my word for this unless someone else backs me up.

 

[unixqueen]

Does this always happen at month end when users are generally busy, as they are now.....is it that your machine is undersized for this time of month?

Your problem could be lots of things, how have you decided that it's not a process issue?....

Have you a memory issue, if you get a memory bottleneck, your system will start swapping and CPU consumption will go up.
This is what I'd do and check for starters....
- If you got it installed use Glance, otherwise install the trial version from the Quarterly apps CD's. Also think about installing Measureware so that you can collect the stats and look at the issue over a period of time.
In the meantime, is #top showing any process culprits
- Find out if your machine is swapping with #swapinfo -ta.
- Is anything being reported to logs, syslog, dmesg
- have you made sure that your Oracle patches are up to date and even checked with your Oracle supplier whether they have seen similar issues with other customers
- Buy some ear plugs (sorry, that's my humourous suggestion attempt)

regards
Clare

 

[bi]

Check your kernel parms and make sure they are high enough. (I know this isn't a comfort, because you'll have to boot the box if you do have to change them. Sorry.)

I have a list of parms recommended by Oracle and can put there here if you want.

Good luck.

 

[mariocq]

Hello:

Have you seen the alert files of Oracle to see if in them appears something interesting?
Check that your file system support largefiles.
Check the syslog.log file to see if in it appear some message related with a lack of resources.
There is a bug related with the at command HPUX and Oracle, do you use it?
I think the first step should be look in the alert files of Oracle.

I hope this help.

 

[SpiritOfLennon]

Firstly I'd like to thank everyone for there advice, we have solved the problem, HP had told us to set some Kernel parameters incorrectly, they very kindly came in and corrected their screw up. I have taken note of all your advice and am reviewing the set up of the system. Thanks again. SOL
I'm only guessing but my guess work generally works for me.