[Bug 53174] New: JIO default maxConnections is not a good default

[bu...@apache.org]

https://issues.apache.org/bugzilla/show_bug.cgi?id=53174

          Priority: P2
            Bug ID: 53174
          Assignee: dev@tomcat.apache.org
           Summary: JIO default maxConnections is not a good default
          Severity: normal
    Classification: Unclassified
          Reporter: fhanik@apache.org
          Hardware: PC
            Status: NEW
           Version: 7.0.27
         Component: Connectors
           Product: Tomcat 7

http://svn.apache.org/viewvc?view=revision&revision=1099855

This feature automatically sets maxConnections to maxThreads. It has a negative
effect for the following reasons

1. It completely changes behavior from Tomcat 6 to 7, and unless one reads the
fine print, one will never find out

2. In servlet 3.0, with the introduction of async servlets, there is a way to
keep a connection open without having a worker thread associated with it.

For case 2. the system SHOULD be able to handle this type of usage through its
default configuration. 

Reverting this fix, and correcting the documentation has no foreseeable impact
on systems that get upgraded with the next release of Tomcat

-- 
You are receiving this mail because:
You are the assignee for the bug.

[Bug 53174] JIO default maxConnections is not a good default

[bu...@apache.org]

https://issues.apache.org/bugzilla/show_bug.cgi?id=53174

Mark Thomas <ma...@apache.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 OS|                            |All

--- Comment #1 from Mark Thomas <ma...@apache.org> ---
Read the following for background:
http://tomcat.markmail.org/thread/wmirgjskvkm4x52k

re point 1 - r1099855 should ensure that the behaviour is the same for both
Tomcat 6 and Tomcat 7.

re point 2 - That is true for write but not true for read. With maxConnections
> maxThreads and blocking IO there is always the risk that a connection with
data to read is unable to acquire a thread while all threads are tied up by
connections without data to read.

Reverting the fix has significant impact since it opens up the risk of problems
as described in the thread from a year ago.

With the increasing move towards async, there is an increasinglg strong
argument for making NIO rather than BIO the default and potentially going as
far as dropping the BIO connector all together but that is a debate that
belongs on the dev list not on this issue.

-- 
You are receiving this mail because:
You are the assignee for the bug.

[Bug 53174] JIO default maxConnections is not a good default

[bu...@apache.org]

https://issues.apache.org/bugzilla/show_bug.cgi?id=53174

Filip Hanik <fh...@apache.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|---                         |WONTFIX

--- Comment #2 from Filip Hanik <fh...@apache.org> ---
(In reply to comment #1)
> 
> With the increasing move towards async, there is an increasinglg strong
> argument for making NIO rather than BIO the default and potentially going as
> far as dropping the BIO connector all together but that is a debate that
> belongs on the dev list not on this issue.

Agreed.

-- 
You are receiving this mail because:
You are the assignee for the bug.