Re: Squid load-balancing and failover, way-prototype

> > Ok, been workin away down here. Warning: this is hacked up code, it is not
> > even beta quality, however it does work (beta functionality, pre-alpha
> > style :) Patch is against squid-2.3.STABLE4
>
> Nice. :-)
>
> A few observations, none about your code:
>
> * Squid's dlin list stuff and mempools are very useful.
> YTou might want to look at using it (you mentioned you were looking
> at using the dlink list stuff, but not mempools, which would
> be nicer ..)

From my megre understanding of mempools, it is code to allow the rapid
allocation of memory from a "pool" of previously allocated, or
pre-allocated memory. Since I only allocate memory at every initialisation
(startup, and reload) I wouldn't see this serving any purpose which is why
I hadn't thought to use it. If I'm wrong about mempools I'd be grateful
for documentation or discussion explaining how my code might benefit :)
 
> * You might want to look at doing this against squid-HEAD. The changes
> shouldn't be that difficult, but I don't know how much of the
> forwarding code has changed between 2.3STABLE4 and 2.4-DEVEL.

I have considered this, unfortunately at this stage I'm actually using
this code in production, so I kinda need a stable base to work off.
 
> Neat work, if I must say so. Well done! :)

Thankyou very much :)

Further to the previous email, the recent spate of testing has run the
code for 24 hours balancing over 3 hosts, handling almost a million real
user hits from a wide variety of browsers, O/Ss etc, and it performed very
very well, with no operational issues and very effective use of CPU.

My congratulations to all of you who have contributed to squid, it took me
basically less than 8 hours to go from knowing nothing about the squid
code to having this up and running effectively.

Richard.
Received on Tue Oct 10 2000 - 08:02:22 MDT