I've made a few changes to the config and am testing this. I've first set the
min-thread-size on the http thread pool to something much higher than the
default of 2. I've set this to 200. From what I could tell, some sort of
cleaner process was going through and periodically killing/interrupting
threads that were not in use to try to get things back to 2. To try to stop
this I've set the min to 200 which should be higher than we need at the
moment. After making the change, the issue was still happening with requests
that were taking more than 15 minutes. I have bumped up the
request-timeout-seconds to 3600s. I'm a bit afraid to completely disable the
timeout behaviour as I do want to have something take care of 'stuck'
requests if we run into any. I'm testing these changes out and will see if
they've improved the situation. I would like to understand why the
interrupting of 1 thread in the thread pool completely kills the logging. I
have found that by running: set
configs.config.server-config.http-service.access-logging-enabled=false set
configs.config.server-config.http-service.access-logging-enabled=true I can
get the logging restarted if this situation happens but we still miss request
when this happens. Thanks.
--
[Message sent by forum member 'arosso']
View Post: http://forums.java.net/node/900505