Hi again, On Fri, May 18, 2012 at 04:10:02PM +0200, Willy Tarreau wrote: > Hi, > > On Fri, May 18, 2012 at 01:05:10PM +0100, tiago ramos wrote: > > Hi, > > > > We are testing dev10. we are having issues with "balance uri depth", > > it doesn't seem to work at all. > > The same configuration is working in dev7 > > balance uri depth 2 > > Did the balance algorithm changed, i couldn't find it in the > > documentation, are we doing anything wrong? > > Just checked and I observed that it is indeed broken. The bad news > is that I have no idea how to fix it, as it's a design bug :-( > > What happens is that we're trying to hash a parameter from the > request at a moment where we don't have the request pointer anymore, > so we hash from the first byte *after* the request. > > I'm now wondering if we wouldn't need two output pointers in the buffers, > one for data we want to send and one for data really sent. This seems > overkill so I'm going back to the blackboard :-/
OK so I have fixed this. It will be in the next daily snapshot which will appear on the site in a few hours. For the record, all content-based hashes were affected, as well as source-binding on a header and http-send-name-header. In fact, all which requires some parsing of a request that has already been scheduled for being forwarded was broken. Regards, Willy
