On Tue, 5 Apr 2005, Merton Campbell Crockett wrote:
The stale data is returned in the instances where NetDB is used to locate the content. The stale content continues to be returned until a request is made to the Squid cache storing the content for fresh content. After looking at the logs, I can identify the system containing the stale content and use client to force retrieval of fresh content.
Do you have a circular peering relation (i.e. two caches peering with each other, or longer cicular chain)?
There is a known condition in cicular peering relations where a stale object can jump among the caches, kicking it's freshness up.
Regards Henrik