On Mon 25 Aug, 1997, David Luyer <luyer@ucs.uwa.edu.au> wrote:
>(and someone else said they have similar sudden problems)
>
>Have you got the dangerous_ports ACL which is default in newer squid
>versions enabled in yours?  Something like this might happen if someone
>requested a URL such as http://localhost:chargen/ (or echo, etc).  I'm not
>completely sure if it would, tho.  It could be excessive sudden large
>in-transit objects, or it could be something else...

I think I've seen this happen with a large never-ending 'server push' object.

I think I commented upon it at the time to squid-users. There were actually two
such connections on one of my servers, and the client had gone away but
unfortunately Squid hadn't yet dropped the connection to the server. It was
still receiving the data at a large rate, and my process had ballooned to
around 400MB (when it was usually around 180MB). I think I was also puzzled at
why Squid didn't *seem* to have started deleting behind.

To see if this is the case, you need to use the cache manager to view the vm
objects. The other giveaway was that the number of vm objects was much much
lower than normal (because of the two very large ones).

James.