No. You'll have 2 load balancers, with a single Virtual IP Address. When one load balancer fails, heartbeat alerts the other load balancer to take over that virtual IP address, keeping traffic flowing.
Thank you, William Attwood System Engineer, Co-Founder Open Box I.T. Solutions, LLC c. 801-634-6479 On Fri, Oct 25, 2013 at 11:27 PM, Kevin <[email protected]> wrote: > I thought about that, but wouldn't that again give us a single point of > failure at the load balancer? > > - Kevin > > On Oct 26, 2013, at 12:22 AM, William Attwood < > [email protected]> wrote: > > Your best bet is to setup a primary, and secondary HAP instance where one > takes over the IP address when the other fails (KeepAlived). You'll want > to keep HAP and your application servers separate, as well. Just have 2 > app servers that HAP load balances between. > > See this: http://www.howtoforge.com/haproxy_loadbalancer_debian_etch > > > > > > Thank you, > William Attwood > System Engineer, Co-Founder > Open Box I.T. Solutions, LLC > c. 801-634-6479 > > > On Fri, Oct 25, 2013 at 11:17 PM, Kevin <[email protected]> wrote: > >> Yes, I have two servers setup almost identically. They both have HAProxy >> and multiple instances of the backend app. >> >> We’re trying to achieve multiple levels of redundancy to handle anything >> from a whole box going down to an app instance crashing. >> >> - Kevin >> >> On Oct 26, 2013, at 12:02 AM, William Attwood < >> [email protected]> wrote: >> >> You're trying to run 2 instances of haproxy, independent of one another, >> with independent application servers as well? >> >> >> >> Thank you, >> William Attwood >> System Engineer, Co-Founder >> Open Box I.T. Solutions, LLC >> c. 801-634-6479 >> >> >> On Fri, Oct 25, 2013 at 10:41 PM, Kevin <[email protected]> wrote: >> >>> I have haproxy successfully running on one server with multiple >>> instances of our app. What I tried to do was bring it up on a second server >>> configured similarly. >>> >>> Right now we are using cookies for session persistence and it works well >>> with a single server. >>> >>> After bringing on the second server and setting up two DNS A records >>> I’ve notice some issues. >>> >>> Some times the browser will apparently end up requesting different >>> things from the two different servers. I didn’t really expect this >>> behavior, but I can see in the developer tools that it is in fact getting >>> cookies from both servers and this is making the app fail. >>> >>> From what I’ve read it seems like I need to peer the two servers so if a >>> request with a cookie comes in on the second server haproxy will forward >>> that request to the correct instance on the first server. >>> >>> Does this make sense, and if so do I need to configure a stick table >>> somehow or am I barking up the wrong tree? >>> >>> Also, it seems that it’s the image requests that typically might end up >>> going to a different server. Sometimes these come through with no cookie, >>> especially if it’s the first time the site is loading. The app framework is >>> apparently setup for image links to use the actual session id in the url to >>> the image. >>> >>> So the link might look like this: >>> >>> >>> https://server2.domain.com/longstringofcharsthatisasessionid/files/%7B9673-7301-0970-2310-9272%7D/background.png >>> >>> Is it possible to read the sessionid cookie of the first response so >>> that when a request comes through with the sessionid in the url I can >>> direct it to a particular instance of the app? >>> >>> >>> Thanks, >>> Kevin >>> >> >> >> >
