You are viewing a plain text version of this content. The canonical link for it is here.
Posted to users@httpd.apache.org by Don Hill <ju...@gmail.com> on 2010/12/28 15:18:13 UTC
[users@httpd] Re: httpd mod_jk cluster
What I really want to know is there a better design that I should use to
gain performance. for example
1.) create multiple HTTPD servers, 2 servers per machine. Each serving 2
tomcats JVM
2.) use load balancer in workers to handle the load balance to the JVM's.
The current configuration is balancing through the vhosts and each vhost has
a worker for a JVM instance.
On Mon, Dec 27, 2010 at 9:05 PM, Don Hill <ju...@gmail.com> wrote:
> Hi.
>
> I am working on a tomcat 5.5 cluster which is using ajp/1.3 and mod_jk and
> trying to determine the best cluster design given the hardware. I have 2
> xeon 2.3 ghz 2 CPU machines with 38GB ram machine. Currently here is the
> config I am using. The TOMCAT and HTTPD servers are on the same physical
> machine.
>
> Each machine is running HTTPD 1.3 with prefork, the MaxClients is 256 due
> compiled in limits. Each machine has 4 virtualhosts running through one
> instance of HTTPD. Two of the VHOSTS are the same app running on 2 Tomcat
> 5.5 with 8GB RAM(configured by customer). The workers are configured to each
> VHOST meaning for each machine there are 4 workers defined and one worker
> is defined for each VHOST. I will try and depict this below. The current
> load balancing is controlled by F5 and manages the load across 2 machines, 4
> VHOST for each app.
>
> Based on this info can someone recommend if this configuration could be
> improved and if so what would you recommend ?
>
> Attached is a pic of the cluster
>
> Thanks for any help on this.
>
>
>
Re: [users@httpd] Re: httpd mod_jk cluster
Posted by Don Hill <ju...@gmail.com>.
Thanks for your input
On Tue, Dec 28, 2010 at 12:02 PM, Jeroen Geilman <je...@adaptr.nl> wrote:
> On 12/28/10 3:18 PM, Don Hill wrote:
>
> What I really want to know is there a better design that I should use to
> gain performance.
>
>
> Umm.. switch to using mod_proxy_ajp, as the apache documentation suggests ?
>
> It offers a binary interface and much improved speed.
>
>
> for example
>
> 1.) create multiple HTTPD servers, 2 servers per machine. Each serving 2
> tomcats JVM
>
>
> Why ? Because we are on 1.3 and the MaxClients is hard defaulted to 256.
> There are 4 JVM's behind one server and I believe that this could be a
> bottleneck, not allowing > 256 requests to be handled at a time.
> Is your tomcat setup not multithreaded ? Sure it is but seems we are
> limited to the number of requests being served by httpd 1.3
>
>
> 2.) use load balancer in workers to handle the load balance to the JVM's.
> The current configuration is balancing through the vhosts and each vhost has
> a worker for a JVM instance.
>
>
> That doesn't really make any sense. You can load balance connections, but
> what does "load balance through vhosts" mean ? ah, We have a F5 load
> balancer doing the balancing against the 4 vhosts for each application
>
>
>
> On Mon, Dec 27, 2010 at 9:05 PM, Don Hill <ju...@gmail.com> wrote:
>
>> Hi.
>>
>> I am working on a tomcat 5.5 cluster which is using ajp/1.3 and mod_jk
>> and trying to determine the best cluster design given the hardware. I have 2
>> xeon 2.3 ghz 2 CPU machines with 38GB ram machine. Currently here is the
>> config I am using. The TOMCAT and HTTPD servers are on the same physical
>> machine.
>>
>> Each machine is running HTTPD 1.3 with prefork,
>>
>
> You're joking.
> Apache 1.3 is EOL. No longer supported. d-e-d-d DEAD.
>
>
> the MaxClients is 256 due compiled in limits. Each machine has 4
>> virtualhosts running through one instance of HTTPD. Two of the VHOSTS are
>> the same app running on 2 Tomcat 5.5 with 8GB RAM(configured by customer).
>> The workers are configured to each VHOST meaning for each machine there are
>> 4 workers defined and one worker is defined for each VHOST. I will try and
>> depict this below. The current load balancing is controlled by F5 and
>> manages the load across 2 machines, 4 VHOST for each app.
>>
>> Based on this info can someone recommend if this configuration could be
>> improved and if so what would you recommend ?
>>
>>
> Shit yes - replace apache by something from this century. 2.2.17 is
> current.
>
> Then proceed to learn all about mod_proxy_balancer, which was made for this
> kind of setup.
>
>
> --
> J.
>
>
Re: [users@httpd] Re: httpd mod_jk cluster
Posted by Jeroen Geilman <je...@adaptr.nl>.
On 12/28/10 3:18 PM, Don Hill wrote:
> What I really want to know is there a better design that I should use
> to gain performance.
Umm.. switch to using mod_proxy_ajp, as the apache documentation suggests ?
It offers a binary interface and much improved speed.
> for example
>
> 1.) create multiple HTTPD servers, 2 servers per machine. Each serving
> 2 tomcats JVM
Why ?
Is your tomcat setup not multithreaded ?
> 2.) use load balancer in workers to handle the load balance to the
> JVM's. The current configuration is balancing through the vhosts and
> each vhost has a worker for a JVM instance.
That doesn't really make any sense. You can load balance connections,
but what does "load balance through vhosts" mean ?
>
> On Mon, Dec 27, 2010 at 9:05 PM, Don Hill <justj2ee@gmail.com
> <ma...@gmail.com>> wrote:
>
> Hi.
>
> I am working on a tomcat 5.5 cluster which is using ajp/1.3 and
> mod_jk and trying to determine the best cluster design given the
> hardware. I have 2 xeon 2.3 ghz 2 CPU machines with 38GB ram
> machine. Currently here is the config I am using. The TOMCAT and
> HTTPD servers are on the same physical machine.
>
> Each machine is running HTTPD 1.3 with prefork,
>
You're joking.
Apache 1.3 is EOL. No longer supported. d-e-d-d DEAD.
> the MaxClients is 256 due compiled in limits. Each machine has 4
> virtualhosts running through one instance of HTTPD. Two of the
> VHOSTS are the same app running on 2 Tomcat 5.5 with 8GB
> RAM(configured by customer). The workers are configured to each
> VHOST meaning for each machine there are 4 workers defined and one
> worker is defined for each VHOST. I will try and depict this
> below. The current load balancing is controlled by F5 and manages
> the load across 2 machines, 4 VHOST for each app.
>
> Based on this info can someone recommend if this configuration
> could be improved and if so what would you recommend ?
>
Shit yes - replace apache by something from this century. 2.2.17 is current.
Then proceed to learn all about mod_proxy_balancer, which was made for
this kind of setup.
--
J.