httpd-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Alexander Stoll <>
Subject Re: [users@httpd] Narrowing Down A Strange Problem
Date Wed, 15 Sep 2004 17:32:56 GMT
David P. Donahue schrieb:

> No error is received for >1250b responses, it just never gets to the
> client.
> I haven't narrowed it down to a specific byte size, but the pattern of
> positive results vs. time-outs seems to hold at approximately 1250
> bytes.  I've made a large number of GET requests and the pattern has
> held.  Even directory listings from Apache itself fit the pattern if
> there are enough files in the directory to bring the generated response
> above the 1250-ish-byte line.
> If you were able to view the site, might I ask your general location
> and/or ISP?

Hi, this seem to be a MTU related problem on the network layer. 
Locations where access fails, should check with a full packet dump of a 
failed request. For some locations the packets go over a link with MTU 
of 1250b and proper PMTU discovery fails, so the TCP session simply 
stucks until timeout...
Maybe you can determine the hop by comparing the traces...
There not really much you can do with different impact:
Generally lower the MTU of your server to 1250b which impacts the 
performance - more packets needed for same payload - or if you are able 
to determine exactly what ranges fail, intercept sessions with these 
sources and force a lower mtu for only these links on your server or 
If there is a braindead ISP on the hop who denies forwarding of "ICMP 
fragmentation needed" packets, contact him and tell him to pull the plug 
and never reconnect until he employed someone who has knowledge of 
TCP/IP networks...

The official User-To-User support forum of the Apache HTTP Server Project.
See <URL:> for more info.
To unsubscribe, e-mail:
   "   from the digest:
For additional commands, e-mail:

View raw message