[erlang-questions] Schedulers go to sleep even with +sfwi 500 option on R16B03

Defnull define.null@REDACTED
Fri Sep 19 16:57:14 CEST 2014


Thank you Scott! 

I added '+scl false' option, 1.5 days have passed and it looks good now. 
Maybe i should also play with '+sub true' because it have similar effect according to docs. Not sure which one will give better performance. Does it also eliminate schedule collapse?

четверг, 18 сентября 2014 г., 8:50:11 UTC+4 пользователь Scott Lystig Fritchie написал:
> Vasily Demidenok <define.null@REDACTED> wrote:
> 
> 
> 
> vd> It helped and activated all schedulers, but only for several hours.
> 
> vd> Then situation repeated.
> 
> 
> 
> One way is to periodically check and use the same hack to reset.  The
> 
> check doesn't come for free, but at least it avoids manual checking and
> 
> manual intervention.  Many thanks to Steve Vinoski for collaboration on
> 
> that code.
> 
> 
> 
> https://gist.github.com/slfritchie/23220d2bf61b08b9a8b9
> 
> 
> 
> Meanwhile, the only way that I know of to avoid scheduler collapse is to
> 
> use both the +sfwi and +scl flags together, e.g. "+sfwi 500 +scl
> 
> false".  The effect on performance of busy systems varies quite a bit:
> 
> you will want to measure your common workloads yourself to see if the
> 
> medicine is worse than the disease.
> 
> 
> 
> Thanks for pointing out the gist about the CouchDB VM version dilemma.
> 
> It's a great summary.
> 
> 
> 
> -Scott
> 
> _______________________________________________
> 
> erlang-questions mailing list
> 
> erlang-questions@REDACTED
> 
> http://erlang.org/mailman/listinfo/erlang-questions


More information about the erlang-questions mailing list