Am 2019-06-21 um 14:33 schrieb Mark Thomas:
On 21/06/2019 13:13, Osipov, Michael wrote:
Folks,
we're migrating off old hosts to new ones, but remain at 8.5.x (.34 to
.42 at the moment) and Java 8. Surprisingly, an empty Tomcat wih just
manager installed takes way too long too start. More precisely, the
Tomcat is fast, deploying manager app is horribly slow:
That sounds like insufficient entropy.
That was my first idea too...
2019-06-21T13:50:35.233 INFORMATION
[deblndw024v.ad001.siemens.net-startStop-3]
org.apache.catalina.startup.HostConfig.deployDirectory Deployment of
web application directory [/var/opt/tomcat-smartld/webapps/manager]
has finished in [47,818] ms
2019-06-21T13:07:40.484 INFORMATION
[deblndw024v.ad001.siemens.net-startStop-1]
org.apache.catalina.startup.HostConfig.deployDirectory Deployment of
web application directory [/var/opt/tomcat-smartld/webapps/manager]
has finished in [126,041] ms
2019-06-21T13:17:35.885 INFORMATION
[deblndw024v.ad001.siemens.net-startStop-1]
org.apache.catalina.startup.HostConfig.deployDirectory Deployment of
web application directory [/var/opt/tomcat-smartld/webapps/manager]
has finished in [80,368] ms
Which class do I need to set to FINEST to narrow down the cause?
org.apache.catalina.startup.HostConfig.level = FINEST
did not help. If nothing helps, I'd debug the Tomcat code.
If it is happening per webapp then I'd take a look at
SessionIdGeneratorBase but it should log a warning if it takes more than
100ms to start.
A series of 3 thread dumps ~5s apart while the app is starting may also
shed some light. I often find doing a diff between them very enlightening.
I have now done the undeploy/redeploy at least 50 times, the line from
SessionIdGeneratorBase never appeared in the log file and the issue
seems to be non-deterministic after a container (HP-UX) restart now.
Restarting the JVM process doesn't make a difference. I have the feeling
that this might be related to the HP-UX container which ran for weeks
now and had to be restarted due to the OpenSSL update.
I will keep on eye on this and re-report when I am able to trace this
properly. I do have the dumps, but they are pointless because the start
takes mostly < 2000 ms now.
Thanks for the pointers,
Michael
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscr...@tomcat.apache.org
For additional commands, e-mail: users-h...@tomcat.apache.org
