[users@httpd] Re: 2.2.13 mod_proxy: backend connection reused long after ttl expired

[Oliver Schoett <ol...@capgemini-sdm.com>]

Oliver Schoett wrote:

> It appears that the Apache 2.2.13 connection timeout (ttl) does not work.

Some more info:  the backend server sends a "TLSv1 Encrypted Alert" 
spontaneously when a connection is idle for approx. 17s.  The error in 
the log file (with a 502 error code sent to the client) occurs when this 
alert happens in response to another request being sent over the 
connection after approx. 17s idle time.  We even set ttl=5 now and still 
see these events.

How is it possible that a backend connection is reused after 17s idle 
time when ttl=5 and smax=0 in the ProxyPass configuration line?  Could 
this be a bug?

The hypothesis that some other processing within Apache or the client 
communication might have taken a long time and thus caused a delay is 
ruled out by the observation that the preceding response on the backend 
connection was transmitted to the client immediately (i. e., 17s ago), 
and the request causing the error was received immediately before the 
error event.

Regards,

Oliver Schoett



---------------------------------------------------------------------
The official User-To-User support forum of the Apache HTTP Server Project.
See <URL:http://httpd.apache.org/userslist.html> for more info.
To unsubscribe, e-mail: users-unsubscribe@httpd.apache.org
   "   from the digest: users-digest-unsubscribe@httpd.apache.org
For additional commands, e-mail: users-help@httpd.apache.org

[users@httpd] SOLVED: 2.2.13 mod_proxy: backend connection reused long after ttl expired

[Oliver Schoett <ol...@capgemini-sdm.com>]

Oliver Schoett wrote:
> Some more info:  the backend server sends a "TLSv1 Encrypted Alert" 
> spontaneously when a connection is idle for approx. 17s.  The error in 
> the log file (with a 502 error code sent to the client) occurs when 
> this alert happens in response to another request being sent over the 
> connection after approx. 17s idle time.  We even set ttl=5 now and 
> still see these events.

The problem appeared on our backend server connections, but was caused 
by a bad client: the client sent HTTP/1.1 requests, but could not 
understand the chunked response that he received.  So he did not know 
when a response was finished and kept waiting for the connection to 
close.  This timed out after 10s, and then the worker process was 
returned into the pool.  During the time that the client was waiting, 
the server connection had already been idle, so that its idle time was 
at least 10s longer than specified with the ttl parameter.

So any time that the worker waits for a client contributes toward the 
idle time of its server connection.

Regards,

Oliver Schoett



---------------------------------------------------------------------
The official User-To-User support forum of the Apache HTTP Server Project.
See <URL:http://httpd.apache.org/userslist.html> for more info.
To unsubscribe, e-mail: users-unsubscribe@httpd.apache.org
   "   from the digest: users-digest-unsubscribe@httpd.apache.org
For additional commands, e-mail: users-help@httpd.apache.org