[SQU] comm_accept and httpAccept errors followed by Connection reset by peer

I'm getting many, really many of this couple of messages in my system:

comm_accept: FD 8: (104) Connection reset by peer
httpAccept: FD 8: accept failure: (104) Connection reset by peer

Sometimes it appears once per second. I've researched a lot but found nothing
regarding this. I've completely no idea of what this could be and therefore
I've no idea of what information you may need. Here's some:

Redhat 6.2, kernel 2.2.18 (getting errors since 2.2.17).
2 cache directories of 6 GB each, 16 x 256 entries.
Activity: around 2200 requests per minute.
2048 FD set, top use near of 1100.

Any help would be very welcome. :)

[]s,
Conrado

--
To unsubscribe, see http://www.squid-cache.org/mailing-lists.html