Good afternoon.
We have a problem with our production tomcat server in that the CPU will
climb after a restart from 1-2% to 100%. The rate of climb corresponds
to the amount of traffic we receive (the more we have, the faster it
climbs). I noticed a couple days ago (by using Lambda Probe), that we
are getting 'rogue' https threads. These are threads that are stuck in
the Service stage for a particular request. I notice that as these
threads become stuck persistently-serviced, the CPU seems to jump about
6% at a time (for each thread). Once we hit 99-100% CPU, we have about
15 of these and we are required to restart tomcat (as it's not
responding with much priority). Lmbda Probe notes that there are no
Current Busy threads, even though it shows these as being Serviced.
Is there a way to get these to time out?
Specifics:
Tomcat 6.0.16
Java 1_5_16
Conf connection snippets:
<Connector port="80" protocol="HTTP/1.1"
connectionTimeout="20000"
enableLookups="false"
maxThreads="100"
minSpareThreads="5"
maxSpareThreads="40"
redirectPort="443" />
<Connector protocol="org.apache.coyote.http11.Http11NioProtocol"
connectionTimeout="60000"
port="443" minSpareThreads="5" maxSpareThreads="15"
enableLookups="false" disableUploadTimeout="false"
acceptCount="100" maxThreads="100"
scheme="https" secure="true" SSLEnabled="true"
keystoreFile="e:\apache\tomcat6\.keystore"
keystorePass="changeit"
clientAuth="false" sslProtocol="TLS"/>
Session timeout is set to 30 minutes in web.xml.
Bob Morman
EMCSA, MCSA
Enterprise Systems Manager
ASM International Headquarters
http://www.asminternational.org
<blocked::http://www.asminternational.org/>
440/338-5151 x5478
The No. 1 reference on metals casting is back with new ways to improve energy
efficiency, productivity and product performance! Read free sample articles
from the all-new ASM Handbook, Volume 15: Casting, and take advantage of
special pre-publication prices before Dec. 15. For more on ASM Handbooks:
http://asmcommunity.asminternational.org/portal/site/www/MatInformation/Handbooks/.