/ Zope / Apsis / Pound Mailing List / Archive / 2008 / 2008-08 / pound error messages

[ << ] [ >> ]

[ Undefined symbol rpl_malloc / "Jim ... ] [ bottleneck in pound / Albert ... ]

pound error messages
Albert <pound(at)alacra.com>
2008-08-21 18:54:06 [ FULL ]
We've been trying to track down network related problems on our end, and 
found that pound is not logging information fully on errors. For 
example, if pound, after selecting a backend from a service pool, has 
trouble communicating with the backend, the request is not logged 
anywhere, except as LOG_NOTICE or LOG_WARNING -- which is the correct 
behavior, if you want to separate informational messages from errors.

The problem, at least for us, is that on an error, pound only logs the 
error condition it encountered (Connnection timed out, error flush 
headers, etc.), but not the request from the user.  This presents 2 
problems:

1. We don't know what the user was requesting, when the error occurred.
2. Our combined logs on our backends contain more responses than the 
information logged in LOG_INFO -- this happens when pound received a 
requests from the client, sends it to the backend, but encounters a 
problems on receiving a response from backend.  Pound will log the error 
in LOG_NOTICE, but the backend logs the response as successful.

I would like to see pound log more information on an error: the client 
IP address, backend IP address (if available), request URI (basically 
LOG_LEVEL=2 information).  Looking at the code, pound only logs such 
information only in handful conditions.  I'd like the information logged 
in all conditions after it has successfully received a request header 
from the client.  Any thoughts, comments?

On a related note, I found couple of minor issues when looking at our 
error logs:

In v2.3.x, pound, when it has problems reading response from the 
backend, logs something like this:
Aug 17 18:21:57 rp2 pound: (b57cfba0) line too long: ....
Aug 17 18:21:57 rp2 pound: (b57cfba0) e500 can't read header
Aug 17 18:21:57 rp2 pound: (b57cfba0) e500 response error read from 
10.10.1.134:80/GET /industry-snapshot/Transportation HTTP/1.0: Success 
(2.759 secs)

Obviously, these are multiple errors caused by a long line.  I'm not 
sure if anything can (should) be done, so it doesn't have "Success" as 
the error string.

In another instance, I have lines such as:

Aug 20 09:53:27 rp2 pound: (b7254ba0) e500 error copy client cont to 
10.10.1.134:80/POST /industry HTTP/1.1: Success (0.007 sec)

but this is not preceded by any other error messages.
Attachments:  
text.html text/html 2708 Bytes

MailBoxer