Connections don't seem to Keep-Alive

From: Chris <rodent@dont-contact.us>
Date: Thu, 27 Apr 2000 17:47:14 -0400

I am running squid in accellerator mode and my objects are caching correctly. However when a client make a request with Connection: Keep-Alive in its request header, the response header (from squid) contains Connection: close. The socket has not been closed at this point, however. If my client makes another request on the same socket, the send appears to work (from a winsock perspective). However, when the client tries to read the response, a socket error (10053) "Software caused a connection abort" is received.*

*The extended info on this error suggests that "the established connection was aborted by the software in your host machine, possibly due to a data transmission timeout or protocol error."

When I check the squid logs (cache.log or access.log), there is no indication that the second request even made it to squid. My guess is that squid is closing my sockets that I have set to "keep-alive". Requests from the same client are coming in at least *every* second, so I would not expect any timeouts to occur.

If squid is removed from the network (ie, direct client<->server connections), keep-alives work fine.
Does Squid support Connection keep-alives in accellerator mode? Is there something special I need to set in the config?

Thanks in advance for any suggestions!
Received on Thu Apr 27 2000 - 15:50:25 MDT

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:53:05 MST