Generic questions - increase performance

[Bastien Latard - MDPI AG <la...@mdpi.com.INVALID>]

Dear Folks, :-)

 From this source 
<https://tomcat.apache.org/tomcat-7.0-doc/config/http.html>, I read:
"Each incoming request requires a thread [...] If still more 
simultaneous requests (more than maxThreads) are received, they are 
stacked up inside the server socket"

I have a couple of generic questions.

1) *How would the increase of maxThreads will behave with RAM usage? 
e.g.: if I increase by 2, would it be twice more?*

2) *What's the defaults values of maxThreads and maxConnections?*
This post <http://stackoverflow.com/a/25765451/5165343> says 
"maxConnections=10,000 and maxThreads=200"

3) Here is my config (/etc/tomcat7/server.xml):
     <Connector port="8080" protocol="HTTP/1.1"
                connectionTimeout="20000"
                URIEncoding="UTF-8"
                redirectPort="8443" />
*Is there a way to kill the request if someone make a big query (e.g.: 
50 seconds), but either close the connection or get a timeout after 5 
seconds?**(or is it default behavior?)
*
Thanks!

Kind regards,
Bastien

Re: Generic questions - increase performance

[Shawn Heisey <ap...@elyograg.org>]

On 4/13/2016 1:50 AM, Bastien Latard - MDPI AG wrote:
> From this source
> <https://tomcat.apache.org/tomcat-7.0-doc/config/http.html>, I read:
> "Each incoming request requires a thread [...] If still more
> simultaneous requests (more than maxThreads) are received, they are
> stacked up inside the server socket"
>
> I have a couple of generic questions.
>
> 1) *How would the increase of maxThreads will behave with RAM usage?
> e.g.: if I increase by 2, would it be twice more?*

Java threads themselves use a *very* small amount of memory.  The memory
used will be determined by what those threads are *doing*.

> 2) *What's the defaults values of maxThreads and maxConnections?*
> This post <http://stackoverflow.com/a/25765451/5165343> says
> "maxConnections=10,000 and maxThreads=200"

The maxThreads setting defaults to 200 in every container I've looked
at.  The Jetty that comes with Solr has this setting increased to 10000
-- so that the limit is effectively removed for typical installations. 
The tomcat documentation says the following about maxConnections:  For
NIO the default is |10000|. For APR/native, the default is |8192|.

> 3) Here is my config (/etc/tomcat7/server.xml):
>     <Connector port="8080" protocol="HTTP/1.1"
>                connectionTimeout="20000"
>                URIEncoding="UTF-8"
>                redirectPort="8443" />
> *Is there a way to kill the request if someone make a big query (e.g.:
> 50 seconds), but either close the connection or get a timeout after 5
> seconds?**(or is it default behavior?)
> *

We strongly recommend using the Jetty that comes with Solr.  This is the
only deployment option that has official support starting with Solr
5.0.  That Jetty has been properly tuned for Solr -- an out-of-the-box
config for Tomcat (or any other container) isn't tuned.

For stopping queries that take too long, Solr has this setting:

https://cwiki.apache.org/confluence/display/solr/Common+Query+Parameters#CommonQueryParameters-ThetimeAllowedParameter

The timeAllowed parameter does not always work -- it depends on which
phase of a query is taking too long.

You can also configure the soTimeout in your client's TCP settings, and
in Tomcat, to kill the connection if it is idle for too long.

There may be other options that might work, these are the ones that I
know about.

Thanks,
Shawn