Op 17-8-2018 om 14:51 schreef Willy Tarreau:
On Fri, Aug 17, 2018 at 01:41:54PM +0200, Olivier Houchard wrote:
That is true, this one is not a bug, but a pessimization, by using the global
update_list which is more costly than the local one.
Patches attached to do as suggest.
Applied, thank you!
Just a little update :)
The '1.9-dev0-48d92ee 2018/07/30'+your initial 2 patches, is running
correctly using as little cpu as can be expected for its little workload
for 4+ days now. I think i can call it 'fix confirmed', as you already
knew ;) , previously the issue would have likely returned in this time
Ill keep it running for a few more days, and switch back to nbthread 3
then.. Till next time ;)