Crossfire Ping Monitoring
OK so I took a little flack in my first post to LET: http://lowendtalk.com/discussion/10800/vps-showdown-waveride-vs-vps-database-vs-iniz-vs-prometeus-vs-domvps
Some of it is because I'm a little racist against OpenVZ. Some of it was justified - and I built upon that.
The criticism that one ping monitor can not take into account routing issues which are out of the hands of the provider is valid and well taken. I responded to that with what I am calling (I get to name it, right?) "Crossfire Ping Monitoring."
The whole process is open-source. Please, scrutinize and post your thoughts for improvement.
You can see my proof-of-concept running at http://foxpa.ws/crossfire/
Nodeping seems to be a four-letter-word around here. Don't ask me - I've never tried it. I use service-level monitoring solutions like Icinga. But it also seems like "host monitors are unreliable!" is a very convenient excuse for providers that offer inferior service.
Crossfire Ping Monitoring should address both the routing problem and the questionable reliability problem because you can set it up to run on any host YOU want. Every host in the pool is constantly monitoring every other host in the pool - from the frontend you can see how any individual VPS appears from each other VPS.
From this, you can extrapolate whether or not there is a routing issue, an overload/DoS issue or a persistent packet loss issue. In the first few hours of running I was able to identify what could be a problem between Iniz and Prometeus that I would have never seen if I was only monitoring from my servers in Montreal (or most elsewhere). You can see a screencap at:
Everything I've done is completely open sourced and open to your scrutiny. I encourage anyone who wants to deploy this system to do so. I also encourage anyone who makes improvements to publish them - god knows there is room for plenty.
That being said, few people can afford 5 or more geographically diverse VPSes for play-time like I can so I would like to extend an offer to VPS providers:
Give me a free VPS with at least 1T Xit 512MB RAM and 10G storage and I will add your VPS to my monitoring pool. You can tell from the code that I have posted that this is a very objective monitoring solution - you get to see how your server responded to each host and they get to see how you responded to them. If they're having connectivity problems it reflects on them even though they see you as down. With a little effort, I can improve the pattern matching so that downed VPSes don't register as seeing your host down. All it would cost you, at worst, is about 1mbit/s 95th.
If there is enough interest I will turn it into "a thing." i.e. vpsmashup.org or something with pretty graphs. Very pretty graphs. For latency and other nice things.
Good night, and good drinking.