Using MPM: Worker To reproduce: 1: make a change to LoadFactor on a balancer member, within the balancer-manager system 2: send a bunch of requests to the load balancer 3: reload the balancer-manager page. It sometimes reverts to the configured default load factors. Conjecture: this might be when the balancer-manager is invoked from a different process from the one in which the original change was made. This is supported by the fact that just a few sporadic requests don't revert the changes, but high-concurrency testing seems to revert them. I haven't tested the impact of changing other parameters within balancer-manager.
Thank you for the bug report. I will look into it, but it sounds as though the manager is updating only its local copy of the info.
I commented out the section of code below the comentary "/* Set to the original configuration */" at "init_balancer_members" on mod_proxy_balancer.c But It seems not be a nice solution...
(In reply to comment #2) > I commented out the section of code below the comentary "/* Set to the original > configuration */" at "init_balancer_members" on mod_proxy_balancer.c > > But It seems not be a nice solution... This patch prevents mod_proxy_balancer from overwriting the status of workers when creating a new child process (http://svn.apache.org/viewcvs?rev=374929&view=rev). Maybe there's some directions there...
Created attachment 20770 [details] mod_proxy_balancer patch
Comment on attachment 20770 [details] mod_proxy_balancer patch This patch prevents mod_proxy_balancer from overwriting the lbfactor, lbset, and lbstatus of workers when creating a new child process.
Fixed in trunk as r574485 (http://svn.apache.org/viewvc?view=rev&revision=574485), backported to 2.2.x as r575708 (http://svn.apache.org/viewvc?view=rev&revision=575708)