Your last suggestion is what are suits call auto-scaling, and the idea of doing that with HAProxy is really appealing and would be a big plus for us.
As a test tomorrow with 1.4.10/stable idea, let's say a pool of 10 servers had two servers with zero weight, and a script watching the unix socket set their weight proportionate to the rest if a) the leastconn queue size was past some threshold or b) too many servers have gone unavailable. What do you think? On Sun, Feb 6, 2011 at 10:18 PM, Willy Tarreau <w...@1wt.eu> wrote: > Hi Joel, > > On Fri, Feb 04, 2011 at 03:04:40PM -0800, Joel Krauska wrote: > > Awesome - This is a great start. > > > > Putting a server in backup or primary mode would be a nice addition. > > Do you have an example of what purpose it would serve ? I'm asking > because it's not very easy to implement with table-based algorithms, > since the size of the table is determined by the GCD of all active > servers' weights. Thus adding a new server will change the size of > the table. > > It's also a feature I've never seen on other products either, which > makes be doubt about its usefulness. > > What I'd like to implement however is the notion of spare servers, > servers which are activated only above a certain load or when too > many servers were lost. Maybe this is in fact what you're looking > for ? > > > Adding new servers would also be nice, but that's likely to be a harder > > project. > > Yes, it was already proposed but there are quite a number of structural > issues. This as simple as file descriptors to check the servers, and > table size for load balancing come to mind. I'd really prefer that we > spend our time working on a hot reconfiguration (which will bring this > benefit) than on border-line features that are supposed to work around > the lack of hot reconf. > > Regards, > Willy > > >
