>It depends if you are using a parent cache.
We aren't.
>If you are getting FTP objects from your parent, then there is
>no way for *your* squid to detect an aborted transfer from the
>parent (unless the FTP server supports the SIZE command and the
>HTTP reply has a content-length header).
>
>But if you don't use a parent cache, then Squid will detect
>aborted FTP transfers between Squid and ftpget, and it will
>purge the aborted objects from the cache.
It doesn't seem to detect aborted transfers - are there certain cases in
FTP where an aborted transfer can't be detected? We are definitely
ending up with incomplete FTP transfers in the cache.
Our FTPs are currently going EXTREMELY slowly due to problems with our
international link, and it is common to go for a few minutes with little
or no traffic on a connection (we are having a lot of incoming packets
dropped).
FTPing without using the cache works slowly, but it does work more often
than when using the cache - it will still time out occasionally though.
Our read_timeout is set at 15 minutes - our international link isn't
quite that bad I hope. client_lifetime is at 200, and it is certainly
giving up long before that length of time is reached.
Brent
Received on Tue May 20 1997 - 21:34:33 MDT
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:35:13 MST