Trac slowness, apache changes to help

[robe]

TemptorSent (Chris Giorgi ) made some changes to apache mpm prefork config

I'm including the change here, since it contains no big secrets

+

+# Set MaxClients based on memory consumption per worker, which appears to be

+# approximately 64MB unshared per instance. At max loading, this 100 workers

+# should consume 6.4GB or less, leaving sufficient memory for other services

+# to avoid hitting swap. This may be decreased if resource contention is

+# observed, or increased if resources allow and connection numbers are high.

 <IfModule mpm_prefork_module>

 <IfModule mpm_prefork_module>

-    StartServers          5

+    StartServers          10

     MinSpareServers       5

     MinSpareServers       5

-    MaxSpareServers      10

-    MaxClients           25

-    MaxRequestsPerChild   5

+    MaxSpareServers      20

+    MaxClients           100

+    MaxRequestsPerChild   200

 </IfModule>

 </IfModule>

I applied the patch to trac.osgeo.org and did an apache reload.

Things seem to go more smoothly after and I didn't see as much cpu spiking in top or ps aux output. Might be wishful thinking on my part or because of the apache reload.

Anyrate will keep this ticket open. Chris has more changes he'd like to put in place to help.

Discussion about all of these issues are on irc:

​http://irclogs.geoapt.com/osgeo-sac/%23osgeo-sac.2018-02-05.log

[robe]

I'm going to close this. don't think we have much to add and much less of an issue now with recent changes.