Regarding normal traffic handling, where the thread count might change over time, I can see your point. But with a 'static' state, such as right after start up with no traffic, how would
(idle_thread_count < min_spare_threads) be triggered ? I would expect the thread count to be unchanging as there are no reasons to shut down active threads - both since (MinSpareThreads < MaxSpareThreads) and (idle_thread_count == min_spare_threads). On Fri, Oct 18, 2013 at 5:19 PM, Jim Jagielski <[email protected]> wrote: > The mojo for this, in both worker and event is: > > (active_thread_count >= ap_daemons_limit * threads_per_child) > > which is only worried about if: > > (idle_thread_count < min_spare_threads) > > On Oct 18, 2013, at 10:22 AM, Thomas Eckert <[email protected]> > wrote: > > > Hey folks, > > > > there's been quite a few issues with the "scoreboard is full, not at > MaxRequestWorkers" error log message. From what I've found all of them > involved some sort of heavy traffic. I've been wondering about the root > cause for this but now I have a reverse proxy showing this message right > after starting up - no traffic involved at all. > > > > I'm getting the message for a simple reverse proxy with about a hundred > virtual hosts which is strange because > > > > StartServers 3 > > ServerLimit 30 > > MaxClients 1500 > > MinSpareThreads 25 > > MaxSpareThreads 75 > > ThreadsPerChild 50 > > MaxRequestsPerChild 0 > > > > doesn't look like it cannot handle 100 hosts without traffic. The vhosts > themselves look like > > > > <VirtualHost 10.8.17.133:80> > > ServerName my_domain_3 > > SSLProxyEngine On > > RequestHeader set X-Forwarded-Proto http > > <Location /> > > ProxyPass balancer://58be3a18b1c6590f106e3536d31d0447/ > lbmethod=bybusyness > > ProxyPassReverse > balancer://58be3a18b1c6590f106e3536d31d0447/ > > Require all granted > > </Location> > > </VirtualHost > > > > which is really basic. There is only one virtual host with a "real" > configuration (setup is part of a scalability test). > > > > Again, there is no traffic involved, it's happening right after start > up. I checked the processes and the start up is clean, meaning there are no > old left-overs that might interfere somehow. > > > > Any hints on how to debug this ? > >
