Re: Sequent PTX Oracle7 stability

Bradley W Howerter (howerter_at_world.std.com) wrote:
> We've come across a few anomilies with oracle 7.0.12 and
> oracle 7.0.13 on a Sequent running ptx v2.0.4.
> (...)

This autumn we've run benchmarks on multiple platforms (SUN, DEC, Sequent, UNISYS, Olivetti, IBM/AIX, NCR, Siemens Nixdorf, Pyramid - just to mention the well-known ones) and encountered some strange behaviour some times to. The benchmark-programs are all designed to exit on a database error _immediately_. I've checked this twice, each SQL-command is followed by a check of sqlca. The programs are all written in C/embedded SQL. Each benchmark starts 50, 125 or 200 processes and runs 1 hour.

Sometimes (I assume one benchmark out of 100) there has happened what I think is a locking problem: Some processes 'hung' that means they never returned from an SQL-statement. Also they haven't consumed any CPU-time. One time all 200 processes hung, while CPU was idle 100%, but no single error message showed up. The only way to get out of that situation was to kill all processes. These situation showed up on different platforms, but only seldom. Too bad I haven't had the time to analyse that problem in detail.

> So we're wondering about Oracle7's stability on Sequent PTX.

So I'm wondering about Oracle 7's stability at all.

Anyone run into such a situation, too ?
Anyone analysed what the problem was ?

-- 
London *            * Berlin           Klaus ter Fehn        
             x                         Neanderstr. 4        
              <ktf_at_bc3.gun.de>         40233 Duesseldorf   
   Paris *                             FRG / Germany       Tel.: +49-211-676331