It turns out that disabling Nagle in Ajp12Interceptor
fixes this problem of lost data. The question is, why?
I can't imagine that Nagle holds back data during a
socket shutdown. Yet, I don't have any other explanation.
I'm not going to commit a change for this until I understand
why this solves the problem, but at least the fix is now
documented for posterity.
Keith
| -----Original Message-----
| From: Keith Wannamaker [mailto:Keith@wannamaker.org]
| Sent: Wednesday, December 12, 2001 2:59 PM
| To: tomcat-dev@jakarta.apache.org
| Subject: ajpv12 socket closure
|
|
| I've been debugging a socket problem in ajpv12 (NT) where bits
| are being written by Tomcat to the socket (back to mod_jk), but
| jk's read fails with a SHUTDOWN. It's as if the bits don't
| quite make it onto the wire, and the socket closure is a hard close.
|
| I thought the point of so_linger was to specify the amount of time
| to try to send the queue bits onto the wire prior to closing the socket.
|
| Wouldn't this indicate that the present value (100) is too small?
|
| The relevant t-d thread is
| http://w6.metronet.com/~wjm/tomcat/2000/Sep/msg00151.html
|
| Thanks for any info,
| Keith
|
| --
| To unsubscribe, e-mail: <mailto:tomcat-dev-unsubscribe@jakarta.apache.org>
| For additional commands, e-mail: <mailto:tomcat-dev-help@jakarta.apache.org>
|
--
To unsubscribe, e-mail: <mailto:tomcat-dev-unsubscribe@jakarta.apache.org>
For additional commands, e-mail: <mailto:tomcat-dev-help@jakarta.apache.org>
|