Fwd: Objects mix-up with 6.1.5/Linux

Previous Topic Next Topic
 
classic Classic list List threaded Threaded
1 message Options
Reply | Threaded
Open this post in threaded view
|

Fwd: Objects mix-up with 6.1.5/Linux

Bill Erickson
Accidently sent directly to Dag rather than the forum...

As an aside, the known bugs causing the type of corruption seen by Thierry are often associated with Epoch GC.  Folks running really old versions of the product can avoid these by simply disabling Epoch GC's.

Regards,
Bill Erickson


From: "William Erickson" <[hidden email]>
To: "Dag Willy Findal-Fossmo" <[hidden email]>
Sent: Thursday, July 12, 2012 8:10:34 AM
Subject: Re: [GemStone-Smalltalk] Objects mix-up with 6.1.5/Linux

GS 6.1.5 is a pretty old version.  I recall that there have been a number of bugs fixed that could cause object corruption like what you're seeing, usually when a live object get's GC'ed and it's OOP recycled to a new object.  Your best bet is to upgrade when you can.

In the meantime you can check for the scope of corruption by doing:

SystemRepository objectAudit

As suggested by Dag.

This is best run on a system with a single session running (other than the GcGem/s), otherwise certain checks are disabled.  For a production environment this usually means making a copy and running it on another machine where there is only the single session running the objectAudit.  Be aware that this can take a long time to run, especially on a large database.

Regards,
Bill Erickson




From: "Dag Willy Findal-Fossmo" <[hidden email]>
To: [hidden email]
Sent: Thursday, July 12, 2012 1:15:59 AM
Subject: Re: [GemStone-Smalltalk] Objects mix-up with 6.1.5/Linux

Hi Thierry,

On 07/12/2012 06:15 AM, Thierry Thelliez wrote:
Hello,

I know we should upgrade this application to a more recent version…

We just encountered a very strange situation where two objects got mixed up with random ones.  One was an internal Gemstone dictionary collision chain index that became an empty array instead of an Integer, the other one was a DateTime that ended up being an instance of a Stream.   In other words, that was not an application bug, but it looked like the oops pointed to the wrong objects and/or page.

I have never seen that before. We fixed/replaced these objects and the database seems fine.  But is there a way that we could check for database consistency? 
I'd probably do a page audit followed by an object audit. You can read about those in the System Admin Guide.

The only recent change we made was to upgrade our Virtualization environment to the most recent version of Proxmox.  

Has anyone experienced such mix-up?



We've probably experienced the same kind of problems, but it's quite a long time ago, so I can't remember the details, but I think there was a restore from backup and logs involved.


Regards,
Dag Findal-Fossmo
www.computas.no


_______________________________________________
GemStone-Smalltalk mailing list
Archive: http://forum.world.st/Gemstone-Customers-f1461796.html





--
------------------------------------------------------------------------
Bill Erickson        VMware/SpringSource/GemStone/Smalltalk Engineering
Software Engineer    15220 NW Greenbrier Parkway #150 Ph: (503)533-3439
[hidden email]     Beaverton OR 97006               Fx: (503)533-3220
------------------------------------------------------------------------



_______________________________________________
GemStone-Smalltalk mailing list
Archive: http://forum.world.st/Gemstone-Customers-f1461796.html