Re: Oracle problems.

Hoff Hoffman wrote:
>
> In article <9uh0g8$7kk$1_at_newsg1.svr.pol.co.uk>, "Leigh G. Bowden" <LGBowden_at_bowdenfamily.fsnet.co.uk> writes:
> :We have an AlphaServer 2100 4/200 with two CPU's and 1GB of memory and
> :plenty of disks - all shadowed.
>
> That's a comparitively old and slow EV4-class Alpha platform, and one
> gigabbyte of main memory is comparatively small by current configuration
> standards.
>
> :As far as I am aware it is using Oracle 7.1.5 on OpenVMS 6.2 AXP.
>
> You will want to confirm this, since this version information will
> be central to most of the subsequent research that is required here.
> Your version of OpenVMS Alpha lacks various of the enhancements around
> sixty-four bit addressing that were added for V7.0, and thus has the
> VAX limitations around stuffing most everything of consequence into
> the S0 and S1 space (2TB) address range -- this probably likely will
> not help with the apparent leak reported here, but I've encountered
> numerous Alpha (pre-V7) systems with the resulting addressing limits.
>
> :It appears to be unreliable. After about three weeks this system will lock
> :up and the console will report "Insufficient memory for operation" type
> :errors
>
> You will want to get the specific error message, not the type of
> message. (No offense is intended.) There are many similar error
> messages, and the specific and full error message text is invaluable
> evidence when determining the details of the particular problem.
> Abbreviations or any other alterations to the reported text should
> be carefully avoided, given the likelyhood these changes will serve
> only to introduce ambigiuity.
>
> :...and will allow nobody to logon even at the console. The only way to
> :get out of this to power cycle it.
>
> This initially appears to be a memory leak, either in OpenVMS or in
> the database software. You will want to apply the mandatory ECO kits
> for OpenVMS Alpha, and you will want to check with Oracle (classic,
> I assume, since that is not an Rdb version) for any database patches,
> and you will want to configure a sufficiently large system crashdump
> file and learn how to force an OpenVMS system crash from the system
> console.

[Quoted] Sounds familiar! Like Hoff said, check for patches - VMS *AND* Oracle! If memory serves (and it frequently doesn't), this may be a known problem with a fix available.

-- 
David J. Dachtera
dba DJE Systems
http://www.djesys.com/

[Quoted] Unofficial Affordable OpenVMS Home Page:
http://www.djesys.com/vms/soho/