Re: Server Gone South

In article <33260025.1996884_at_news.earthlink.net>,   beverett_at_usa.net (Brian Everett) wrote:
>
>
> I go this in my email today...
>
> <<<" Ran into a problem with the following configuration:
>
> Sun SPARC 20, 2 CPU's
> 256 MB memory
> Sun Solaris 2.5.1
>
> Oracle 7.3.2.3.0
>
> This application has been running fine for 8 months on a box with only
> 128 MB of memory. They gave me the new box with more memory and I
> installed Oracle and migrated the DB (small 100 MB) to the new box.
> We turned in on this morning and for two hours the application
> screamed but then everything turned south. Oracle shut itself down,
> we rebooted, same results. We had to roll the database back to the
> old/smaller server.
>
> The error that is in all the trace and log files is:
>
> ORA-07445 : exception encountered: Core Dump [] [SIGSEGV] ...
>
> We rolled the application back to the smaller box and it is running
> fine. The only variables on the new box are newer version of the OS
> and Oracle 7.3.2.3.">>>
>
> Any ideas would be most appreciated.
>
> Good Luck! Brian

A core dump is a memory problem. I assume that you're using the correct Oracle port for your O/S version - if you aren't you could definitely see something like this resulting from problems with threading, etc. Are you using asynch I/O? You might have threads stepping on each other's critical memory regions. I would try turning off asynch I/O to see if the problem goes away.

When you rebooted, did the db dump core immediately, or did it run OK for a couple of hours and then bomb? If it ran OK for awhile and them bombed, it sounds to me like a thread problem. Either Oracle or Solaris (or the combination) is not completely thread-safe.

It also wouldn't hurt to run hardware tests on the memory.

Good Luck!

Brenda Muller

-------------------==== Posted via Deja News ====-----------------------
      http://www.dejanews.com/     Search, Read, Post to Usenet