On Thu, Apr 14, 2022 at 4:00 PM Ruediger Pluem <rpl...@apache.org> wrote: > > On 4/14/22 3:43 PM, Yann Ylavic wrote: > > > > I have not tested it yet but possibly without this if many children > > segfault successively we might enter a busy fork() loop.
Tested and it seems to work.. > > The above would restart killed children immediately (using the > > remaining_children_to_start mechanism) only if there are less than 3 > > successively, otherwise sleep 1s and > > perform_idle_server_maintenance(). > > > > WDYT? > > Having a stopper for the fork loop sounds sensible. We may should log > something if we hit it to ease analysis for situations where > the childs quickly segfault. r1899858 Thanks; Yann.
