On Sat, 31 Dec 2016 20:10:05 -0700, Jack J. Woehr wrote: > >We have empirical proof of that at one of my clients . Took down a monitoring >system this evening. > >Was peacefully spawning services and suddenly the last-started time was in the >future and it spawned 100 in a second and >choked. > Ah! You could do it with z/OS. It just takes some truly bad coding: When each "service" starts, it fetches CVTLSO; does STCK; adjusts (as TIME macro does) and logs that as the started time.
The "monitor" also fetches CVTLSO when it starts and continues to use that frozen value to adjust STCK. So it thinks it's way behind, and each service it starts is still almost a second tardy. It's the cost of the false economy of moving a double-word fetch out of a not-very-tight loop. That's why C invented the "volatile" storage class, but the programmer forgot to declare CVTLSO volatile. -- gil ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
