On Tue, Feb 6, 2018 at 9:48 AM, Yann Ylavic <ylavic....@gmail.com> wrote: > On Tue, Feb 6, 2018 at 3:33 PM, Eric Covener <cove...@gmail.com> wrote: >> On Tue, Feb 6, 2018 at 9:23 AM, Yann Ylavic <ylavic....@gmail.com> wrote: >>> With my proposal, this is lost only if IP:port(s) or >>> ServerName/Alias(s) change, which is already a win and shouldn't >>> change the way each balancer is bound to its vhost (i.e. a request on >>> a vhost wouldn't be accounted/handled by a lb on another vhost). >> >> I am thinking no ServerAlias in the hash. > > Looks reasonable because taking ServerAlias into account only serves > corner/weird cases (mis)configurations where two vhosts would have the > same everything else... > But if a request asks for the ServerAlias which is only in one of the > two, all bets would be off on the lb side (the SHMs would be reused > but not necessarily with the same size). > This is probably a case which we should error/warn about at > (re)startup anyway, or maybe include ServerAlias in lb only if it > happens?
I see what you mean now, I was only thinking "find the wrong vhost-keyed thing" but not that offsets/etc could be living in real vhost server config and not line up correctly. Not so sure now. -- Eric Covener cove...@gmail.com
