I searched for "monitor" in the archives and read back about a year, and didn't see this asked. I did see
Apologies if it's something that has already been asked and answered. In my application it would be useful to have multiple monitors, because I'd like to use HAProxy as a poor man's network monitor, since it already knows the state of each of its backends. So, for example, if I had three backends--let's call them "static", "app", and "service"--I'd like to be able to define: /monitor/alive /monitor/healthy Those would be true if, in the first case, at least one instance of each backend was OK, and in the second case, if all instances of all backends were OK. Then I'd also like: /monitor/static/alive /monitor/static/healthy and so on for app and service. *THEN* I could simply iterate through my list of monitor endpoints and quickly build a graph of the system's overall health. As it stands, though, you get one monitor URI per frontend. It looks like it would take some surgery to turn that from a simple string to a hash associating names with particular monitors, and then a separate pointer per name referring to an ACL to do the monitor fail condition. So I'm wondering the following: 1) is this a lot harder than it looks? Obviously there's a bit more computation to do the "is this URI a monitor-uri hash key? If, look up the associated ACL" than the strcmp it currently is. That doesn't feel like it should be a huge performance penalty, but, well, a C hash is a lot heavier than a strcmp. Even so, strmap (http://pokristensson.com/strmap.html) (which requires both keys and values to be strings, so it's not a totally generic hash implementation) would be perfectly OK for this purpose, and it's LGPL, so should be safe to integrate. 2) Am I thinking about this the wrong way? That is, is the thing I want to do just silly and I should be going about it some other way? 3) Am I really the first person to make this feature request? It seems like an obvious thing to want to do. I seem to recall at some point finding some discussion that you could do this by chaining haproxies, but that seems much heavier-weight than just implementing: monitor-hash app-healthy /monitor/app/healthy monitor-fail-hash app-healthy nbserv(app) lt 4 Thanks, Adam
