On Sat, May 10, 2014 at 2:18 AM, C Anthony Risinger <[email protected]> wrote:
> On Fri, May 2, 2014 at 4:23 PM, Matt Phipps <[email protected]>wrote: > >> On Fri, May 2, 2014 at 3:59 PM, Roberto De Ioris <[email protected]> >> wrote: >> > --reload-on-exception seems the best approach to me. >> >> Yep, that seems like the way forward. > > > also consider some form of: > > --thunder-lock > --ignore-write-errors > --disable-write-exception > --write-errors-exception-only > --write-errors-tolerance > > ...at some point last year we added --reload-on-exception as a band-aid > for <insert-reason>, and left it since the option seemed a sane default > anyway. > > BUT. > > for various reasons our app has a sluggish boot up time, not really > significant just slightly annoying, say 3-5 seconds. > ...AND NOW FOR THE EPIC CONCLUSION. these per-process 3-5 seconds restart times in conjunction with a peak traffic time and latency-tuned-with-excellent-record --cheaper-[busyness]-* options combined with user fervent (ie. mashing reload after N seconds == raising a write error and killing worker) in such a way that effectively achieved total lockout -- processes reloading only to be killed moments later. since some of the boot time is spent preloading modules and precaching <insert-here>, the databases were grinding away in circles at 9-11+ loadavg.. JFYI... fun times :) -- C Anthony
_______________________________________________ uWSGI mailing list [email protected] http://lists.unbit.it/cgi-bin/mailman/listinfo/uwsgi
