You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@geronimo.apache.org by Łukasz Budnik <lu...@gmail.com> on 2010/05/26 15:03:24 UTC
very, very, very slow start of tomcat6 car
Hi all,
I have been migrating my application from G 2.1.3 to G 2.1.4 and today
I have finally migrated the production environment.
But I have one pretty annoying issue.
In dev both G2.1.3 and G2.1.4 start Tomcat module in a reasonable time.
But in production the start time is always ~2 minutes! For example:
xh@apl05:/opt/geronimo-tomcat6-javaee5-2.1.4/bin# tail -f
/opt/geronimo-tomcat6-javaee5-2.1.4/var/log/geronimo.out
Module 16/67 org.apache.geronimo.configs/activemq-broker/2.1.4/car
started in .601s
Module 17/67 org.apache.geronimo.configs/openjpa/2.1.4/car
started in .000s
Module 18/67 org.apache.geronimo.configs/webservices-common/2.1.4/car
started in .000s
Module 19/67 org.apache.geronimo.configs/openejb/2.1.4/car
started in .250s
Module 20/67 org.apache.geronimo.configs/axis/2.1.4/car
started in .050s
Module 21/67 org.apache.geronimo.configs/axis2/2.1.4/car
started in .000s
Module 22/67 org.apache.geronimo.configs/axis2-ejb/2.1.4/car
started in .000s
Module 23/67 org.apache.geronimo.configs/j2ee-corba-yoko/2.1.4/car
started in .452s
Module 24/67 org.apache.geronimo.configs/jasper/2.1.4/car
started in .000s
Module 25/67 org.apache.geronimo.configs/tomcat6/2.1.4/car
started in 2:34.440s
Module 26/67 org.apache.geronimo.configs/tomcat6-no-ha/2.1.4/car
started in .000s
2:34.440s? Give me a break...
Just for comparison:
My dev machine is a laptop with Windows 3 GB and Intel Core Duo @ 2.20GHz.
In production I have Linux 64 bits with 16 GB and 8 x Intel Xeon @ 2.50GHz.
Am I the only one experiencing this kind of issue?
cheers,
Łukasz
Re: very, very, very slow start of tomcat6 car
Posted by Łukasz Budnik <lu...@gmail.com>.
Hi Kevan,
You were right! It was entropy.
Now it looks much, much better:
Module 25/68 org.apache.geronimo.configs/tomcat6/2.1.4/car
started in .851s
thanks!
Łukasz
On 26 May 2010 15:40, Kevan Miller <ke...@gmail.com> wrote:
>
> On May 26, 2010, at 9:03 AM, Łukasz Budnik wrote:
>
>> Hi all,
>>
>> I have been migrating my application from G 2.1.3 to G 2.1.4 and today
>> I have finally migrated the production environment.
>>
>> But I have one pretty annoying issue.
>>
>> In dev both G2.1.3 and G2.1.4 start Tomcat module in a reasonable time.
>>
>> But in production the start time is always ~2 minutes! For example:
>>
>> xh@apl05:/opt/geronimo-tomcat6-javaee5-2.1.4/bin# tail -f
>> /opt/geronimo-tomcat6-javaee5-2.1.4/var/log/geronimo.out
>> Module 16/67 org.apache.geronimo.configs/activemq-broker/2.1.4/car
>> started in .601s
>> Module 17/67 org.apache.geronimo.configs/openjpa/2.1.4/car
>> started in .000s
>> Module 18/67 org.apache.geronimo.configs/webservices-common/2.1.4/car
>> started in .000s
>> Module 19/67 org.apache.geronimo.configs/openejb/2.1.4/car
>> started in .250s
>> Module 20/67 org.apache.geronimo.configs/axis/2.1.4/car
>> started in .050s
>> Module 21/67 org.apache.geronimo.configs/axis2/2.1.4/car
>> started in .000s
>> Module 22/67 org.apache.geronimo.configs/axis2-ejb/2.1.4/car
>> started in .000s
>> Module 23/67 org.apache.geronimo.configs/j2ee-corba-yoko/2.1.4/car
>> started in .452s
>> Module 24/67 org.apache.geronimo.configs/jasper/2.1.4/car
>> started in .000s
>> Module 25/67 org.apache.geronimo.configs/tomcat6/2.1.4/car
>> started in 2:34.440s
>> Module 26/67 org.apache.geronimo.configs/tomcat6-no-ha/2.1.4/car
>> started in .000s
>>
>> 2:34.440s? Give me a break...
>>
>> Just for comparison:
>>
>> My dev machine is a laptop with Windows 3 GB and Intel Core Duo @ 2.20GHz.
>>
>> In production I have Linux 64 bits with 16 GB and 8 x Intel Xeon @ 2.50GHz.
>>
>> Am I the only one experiencing this kind of issue?
>
> Hi Lukasz
> I'll bet you a reasonably-priced-beverage of your choice that your Linux production environment is running low on System entropy. The long delay is occurring while pseudo-random numbers are being generated for the Tomcat SSL Connector.
>
> A 'kill -3 <geronimo-process-id>' during your 2 minute wait should generate thread stack traces that would tell us for sure.
>
> Searching the geronimo user list for 'entropy' should find some previous discussions on the subject. There are some tricks for making the random number generator slightly less random, but less likely to encounter an entropy issue.
>
> --kevan
>
>
Re: very, very, very slow start of tomcat6 car
Posted by Kevan Miller <ke...@gmail.com>.
On May 26, 2010, at 9:03 AM, Łukasz Budnik wrote:
> Hi all,
>
> I have been migrating my application from G 2.1.3 to G 2.1.4 and today
> I have finally migrated the production environment.
>
> But I have one pretty annoying issue.
>
> In dev both G2.1.3 and G2.1.4 start Tomcat module in a reasonable time.
>
> But in production the start time is always ~2 minutes! For example:
>
> xh@apl05:/opt/geronimo-tomcat6-javaee5-2.1.4/bin# tail -f
> /opt/geronimo-tomcat6-javaee5-2.1.4/var/log/geronimo.out
> Module 16/67 org.apache.geronimo.configs/activemq-broker/2.1.4/car
> started in .601s
> Module 17/67 org.apache.geronimo.configs/openjpa/2.1.4/car
> started in .000s
> Module 18/67 org.apache.geronimo.configs/webservices-common/2.1.4/car
> started in .000s
> Module 19/67 org.apache.geronimo.configs/openejb/2.1.4/car
> started in .250s
> Module 20/67 org.apache.geronimo.configs/axis/2.1.4/car
> started in .050s
> Module 21/67 org.apache.geronimo.configs/axis2/2.1.4/car
> started in .000s
> Module 22/67 org.apache.geronimo.configs/axis2-ejb/2.1.4/car
> started in .000s
> Module 23/67 org.apache.geronimo.configs/j2ee-corba-yoko/2.1.4/car
> started in .452s
> Module 24/67 org.apache.geronimo.configs/jasper/2.1.4/car
> started in .000s
> Module 25/67 org.apache.geronimo.configs/tomcat6/2.1.4/car
> started in 2:34.440s
> Module 26/67 org.apache.geronimo.configs/tomcat6-no-ha/2.1.4/car
> started in .000s
>
> 2:34.440s? Give me a break...
>
> Just for comparison:
>
> My dev machine is a laptop with Windows 3 GB and Intel Core Duo @ 2.20GHz.
>
> In production I have Linux 64 bits with 16 GB and 8 x Intel Xeon @ 2.50GHz.
>
> Am I the only one experiencing this kind of issue?
Hi Lukasz
I'll bet you a reasonably-priced-beverage of your choice that your Linux production environment is running low on System entropy. The long delay is occurring while pseudo-random numbers are being generated for the Tomcat SSL Connector.
A 'kill -3 <geronimo-process-id>' during your 2 minute wait should generate thread stack traces that would tell us for sure.
Searching the geronimo user list for 'entropy' should find some previous discussions on the subject. There are some tricks for making the random number generator slightly less random, but less likely to encounter an entropy issue.
--kevan