We looked at a few other solutions like ServerDensity etc.
The main issue was that we wanted to report across all requests, e.g. there is no single health check endpoint that an external system can hit to determine if everything is OK or not.
The crux of the matter is that we needed to parse our Nginx logs for the response code on every single request and calculate the uptime from that.
I may be missing something, but checking your logs does not fulfill exactly the same function as an external check.
I mean, if X% of your requests never make it to the logs cause ELB is messed up/dns is fucked up/disk is full/nginx dies/grasshoppers then you'd happily show that everything is ok, since no new 500 status gets pulled from the logs.
So internal and external checks should be compounded, imo.
The main issue was that we wanted to report across all requests, e.g. there is no single health check endpoint that an external system can hit to determine if everything is OK or not.
The crux of the matter is that we needed to parse our Nginx logs for the response code on every single request and calculate the uptime from that.