Re: Feature request

From: Robert Collins <robertc@dont-contact.us>
Date: 02 Nov 2002 19:36:29 +1100

On Sat, 2002-11-02 at 13:59, Ray Cole wrote:
> I have a usleep-free flavor of this now. I found when I removed the
usleep that it actually started misbehaving again. I was only modifying
neighbor.c, when really I also needed to make a similar modification to
forward.c. With the usleep in place it was giving enough time for
forward.c's retry attempts (which is set at 10) to recover - more or
less masking the issue.
>
> So I do have this going without any usleep. The configuration
parameter per peer sets the amount of time to retry connections when
they fail - default is 0 to indicate this isn't wanted. It will
basically ignore n_tries on FwdState if this is set, retrying up to X
seconds. Generally speaking, 1 second is more than sufficient - almost
makes me want to make it just an on/off switch but conceivably someone
might want just a little longer.

Sounds much better - excellent.

> I'm still not quite pleased with the way I'm implementing it, but at
least now I have a definitive answer on the area of code that needs the
improvements. I'd also rather make this a little more generic - perhaps
an option when starting a connection in comm.c rather than only
impacting peer connections (perhaps configurable at the peer level or
globally, just in case someone experiences this problem in other areas
aside from peers).

Hmm. I'd need to review comm.cc myself, but IIRC failures are handled by
the comm client code, not by comm itself (as it should be), so
forward.cc is the right place to do it. (Forward forwards to both peers
and origin servers).

> I won't put anything away yet. Since this is my first time to
contribute I'd rather take my time to make sure I'm not going to screw
something up.... :-) I'm also only modifying 2.5-STABLE1 on my machine
at this time. If anyone wants to try it with 2.5 let me know, otherwise
I'll keep it to myself for a few days.

Thats cool, I'm really only hacking on 2.5 when fixing bugs :}.
 
> BTW - should I subscribe to the squid-dev list at this point, or
should I remain unsubscribed?

You are welcome to join. The list volume varies, but typically only 5-10
messages a day.

> I also wanted to say that I was a bit hesistant to contribute at
first, but everyone seems to have a good attitude about such things. I
wasn't sure what I was going to get myself into - so thanks! :-)

You're welcome :}.

Cheers,
Rob

Received on Sat Nov 02 2002 - 01:36:32 MST

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:18:36 MST