Sibling hits infinitely delayed.

I've noticed that if one of our cluster of caches goes down, that sibling
queries to it don't appear to be timing out.

Let me clarify.

Clearly the ICP queries stop getting sent, since it is no longer
responding. However if there is an actual fetch (resulting from an ICP
hit) going on, and the server goes away, it doesn't appear to time out.

Worst examples from the logs:

897509399.028 11999139 195.153.175.9 ERR_CLIENT_ABORT/000 0 GET http://www.zdnet.com/anchordesk/images/sp_dah99CCFF.gif - SIBLING_HIT/haar.pipex.net -
897509419.047 11999890 193.149.91.44 ERR_CLIENT_ABORT/000 0 GET http://zone.com/Graphics/Scheme03/pgidzlist.gif - SIBLING_HIT/haar.pipex.net -
897509425.037 11999646 193.149.91.44 ERR_CLIENT_ABORT/000 0 GET http://zone.com/graphics/scheme03/imgprem.gif - SIBLING_HIT/haar.pipex.net -

This is squid 1.1.21.

John
Received on Thu Jun 11 1998 - 03:22:47 MDT