Re: multiple A records

From: Michael Pelletier <mikep@dont-contact.us>
Date: Tue, 1 Jul 1997 13:16:49 -0400 (EDT)

--MimeMultipartBoundary
Content-Type: TEXT/PLAIN; charset=US-ASCII

On Tue, 1 Jul 1997, Oskar Pearson wrote:
> time-spent would put a limit on the time that it tries to connect - people
> will get irritated if it takes 20 minutes for it to tell them "sorry, I can't
> connect"......

I handle this by trying to cut back the connect timeout down to a minimum
of 5 seconds, depending on how many addresses are available. If a server's
just busy, it'll return a connection refused almost immediately, though.

> > **Probably the main reason for most coredumps in Squid is when
> > we issue a callback and the 'callback_data' memory has been freed.
> > This is why we have things like ipcache_unregister() etc. But
> > note there is no "connect_unregister()."
>
> how does Michael Pelletier's patch handle this? or does it rely on
> his malloc not "truly" giving it back to the OS and therefore still
> having the right values until by chance we next xmalloc it?

My patch completely ignores this little contingency. My mind isn't quite
that far around the Squid code just yet... ;-) This could be why it
doesn't stay up for weeks at a time like a normal 1.1.10 would. Although
until today, it had been up for about a week with fairly steady usage.

        -Michael Pelletier.

--MimeMultipartBoundary--
Received on Tue Jul 29 2003 - 13:15:42 MDT

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:11:20 MST