RE: Strange Elapsed In Raw Trace
- From: "Larry Elkins" <elkinsl@xxxxxxxxx>
- To: <cary.millsap@xxxxxxxxxx>, "'Oracle-L (E-mail)'" <oracle-l@xxxxxxxxxxxxx>
- Date: Thu, 14 Oct 2004 19:11:15 -0500
Thanks. I'll check out the NTP angle.
Regards,
Larry G. Elkins
elkinsl@xxxxxxxxx
214.954.1781
> -----Original Message-----
> From: oracle-l-bounce@xxxxxxxxxxxxx
> [mailto:oracle-l-bounce@xxxxxxxxxxxxx]On Behalf Of Cary Millsap
> Sent: Wednesday, October 13, 2004 9:31 PM
> To: 'Oracle-L (E-mail)'
> Subject: FW: Strange Elapsed In Raw Trace
>
>
> Oracle tim values can wrap. But I would expect that seeing that event =
> would
> be exceptionally rare.
>
> The original posting looks like something Mike Brown of Glen Raven just
> discussed with me last week in Charlotte. If your db server is an NTP
> client, then it's possible for an e or ela value to look like a huge
> positive integer on platforms where the Oracle porters haven't been =
> careful
> in regarding negative numbers as actual negative numbers. For example,
> consider the following sequence of events:
>
> e0=3D gettimeofday returns n seconds since Epoch
> Oracle dbcall processing takes place
> NTP adjustment takes place, setting the clock back k seconds
> More Oracle dbcall processing takes place
> e1 =3D gettimeofday returns n - k seconds since Epoch (k>0)
> Oracle calculates e =3D e1 - e0 =3D (n - k) - n =3D -k
> Oracle mistakenly emits e=3D-k in unsigned integer format
> ...which looks like a colossal positive e value
>
> Happens at least on Oracle for HP-UX.
>
>
> Cary Millsap
> Hotsos Enterprises, Ltd.
--
//www.freelists.org/webpage/oracle-l
Other related posts: