Re: How to resolve core dumps from ORA-07445: exception encountered: [audfro()+4976]
Date: Sun, 7 Jun 2020 16:24:02 -0400
Message-ID: <b54ef5f4-374b-3dc7-441d-5e068905a9e9_at_gmail.com>
Hi Dave!
ORA-07445 happens when the OS, usually the paging system, sends the SIGSEGV, signal 11 by the POSIX standard, to an Oracle process. The occurrence is usually a consequence of following the NULL pointer. Hitting the address space limits usually causes ORA-04035. That means that ORA-07445 signifies a bug in the Oracle code. The only way to deal with Oracle bugs is to call the Oracle support. They will either tell you that the bug is already known and give you the patch to install or they will thank you for discovering the bug and create a patch for you, depending on the priority. What I don't know is how this is implemented on Winduows, which are not POSIX compliant and don't have the POSIX signalling mechanism. Solaris is POSIX compliant, so you're good there.
Regards
On 6/7/20 1:05 PM, Herring, Dave (Redacted sender HerringD for DMARC) wrote:
>
> Charles,
>
> In production where this happens, is the sequence of events that cause
> the core dumps consistent, in that can predict what will generate this
> specific core dump? I'm wondering if you can find out a way to avoid
> the issue causing the core dump. We've had issues in the past where
> errors like this were generated by application users who, in turns
> out, where performing operations that weren't necessary so instead of
> patching/upgrading we were able to stop the behavior.
>
> Regards,
>
> Dave
>
> *From:* oracle-l-bounce_at_freelists.org <oracle-l-bounce_at_freelists.org>
> *On Behalf Of *Charles Schultz
> *Sent:* Wednesday, June 3, 2020 7:27 AM
> *To:* ORACLE-L <oracle-l_at_freelists.org>
> *Subject:* How to resolve core dumps from ORA-07445: exception
> encountered: [audfro()+4976]
>
> *CAUTION:*This email originated from outside of D&B. Please do not
> click links or open attachments unless you recognize the sender and
> know the content is safe.
>
> Good day!
>
> We have been receiving, on average, about 30 core dumps per day in one
> of our production databases:
>
> ORA-07445: exception encountered: core dump [audfro()+4976] [SIGSEGV]
> [ADDR:0x49D594984] [PC:0x10C82C930] [Application data integrity
> precise] [2]
>
> We have an open SR with Oracle Support for the past 5 months. We _had_
> a work around (disable ORA_SECURECONFIG and ORA_LOGON_FAILURES) that
> put a stop to the core dumps; however, as of the January 2020 RU that
> workaround no longer works and we are again getting core dumps (even
> though no audit policies are currently enabled). Due to several
> factors, we are unable to apply any more recent RU until August. At
> this point, we do not have enough business justification to roll back
> the January RU, as the core dumps are not yet impacting business
> operations (they do fill up the filesystem, which we are currently
> managing with cron jobs).
>
> We are seeking help and expertise in how to prevent further core
> dumps. Anyone have any bright ideas?
>
> * Oracle EE 12.2.0.1 (now with the January 2020 RU)
> * Solaris 11
> * The issue does not reproduce in non-Prod, but we do have one other
> production database that has similar issues (however, it has over
> 850+ enabled audit policies).
> * We are currently using pure unified auditing
> * No RAC, no tenant databases (not CDB/PDB)
> * no enabled unified audit policies
>
> Bugs found by Oracle Support as being possibly related:
>
> * Bug 29848736 - ORA-07445 [AUDFRO()+3128] [APPLICATION DATA
> INTEGRITY PRECISE] WITH UNIFIEDAUDIT
> * Bug 24316947 - ORA 7445 AND ORA 600 AFTER APPLYING
> 11204/12101/12102 APR DBPSU/DBBP
> * Bug 30264084 - SIGSEGV, APPLICATION DATA INTEGRITY PRECISE]
> [ADDR:0X3C5750B84] [PC:0X10C765A64, AUDFRO()+3172]
> * Bug 29970261 - XF20.1SEC_UNIAUD_FGA - TRC - KZAPSYSOPTCHK -
> ORA-7445 [KZAPSYSOPTCHK()]
>
> --
>
> Charles Schultz
>
-- Mladen Gogala Database Consultant Tel: (347) 321-1217 -- http://www.freelists.org/webpage/oracle-lReceived on Sun Jun 07 2020 - 22:24:02 CEST