I'm Kiran Gollu, one of the founders of neptune.io. Currently, we support AWS. However, our SaaS is hosting platform agnostic, and our agent offering can be easily adapted to private data centers. We will add support for private data centers in the future versions.
@jmarrapodi, great to hear!. Every server on an average generates about 5 alerts per month and if they are not handled properly, they will cause outages. In fact, amazon (AWS) and facebook (https://www.facebook.com/note.php?note_id=10150275248698920) built internal tools to handle this exact problem. Our mission is to democratize it for everyone!
This was based on empirical survey with over 80 companies (startups, medium size and big enterprises), and this is also in line with our personal experience managing thousands of servers at AWS.
The thing about averages is it hides the good and the bad into one lump number. If your systems were forced to rapidly scale up, maintain various features, and need to refactored/rearchitecured, your probably seeing more than 5 alerts a month. Which makes something like neptune.io usefull since customizing the response based on alerts is your only choice in the short term. Not many if at all any services out there offer solutions to poorly functioning systems. And just to add, at the pace some startups grow, you'll never have a perfect set of systems in place, so you just deal with things as they come. Great systems are a luxury for startups.
I had a great experience working with the neptune.io service to automate some of the remediation tasks that I used to run manually on my cloud servers. Keep it up guys. Looking forward to many more powerful and cool features.
Great concept! I'm in the process of setting up a Neptune account, and already I can see a couple clear use cases:
* For alarm conditions that don't have a scriptable fix, but DO generally entail the same sequence of diagnostic steps (checking running processes, memory usage, etc.), set up Neptune to automatically run these diagnostics and email the output to our PagerDuty team. Then, by the time they log into a machine, they already have all the information they need to act.
* For alarm conditions that DO have a scriptable fix, instead of automatically running the script, email the script out to our entire ops team. This way, even if the most expert person isn't available immediately, someone else may have the information they need to fix the situation right away.
Besides CloudWatch, what are some of the other integrations you're looking to implement next?
It will be interesting to know what kind of integration is planned for Nagios as it already supports Event handlers[1] for executing script when there's a state change.
@Ecio, we are different in a few ways:
1) we are monitoring tool agnostic
2) we provide a much better UX than clunky nagios interface
3) unlike event handlers, we provide best practice script templates so that you don't have to figure out entire script yourself
4) finally, we offer rate limiting features for executing scripts (think don't run this action more than 3 times in last one hour).
In a way, neptune is your virtual first-level on call engineer. If we can resolve alerts for you automatically, neptune won't bother you. If action fails for some reason, it's escalated automatically to right on call engineer.
12 comments
[ 172 ms ] story [ 1458 ms ] threadSource?
* For alarm conditions that don't have a scriptable fix, but DO generally entail the same sequence of diagnostic steps (checking running processes, memory usage, etc.), set up Neptune to automatically run these diagnostics and email the output to our PagerDuty team. Then, by the time they log into a machine, they already have all the information they need to act. * For alarm conditions that DO have a scriptable fix, instead of automatically running the script, email the script out to our entire ops team. This way, even if the most expert person isn't available immediately, someone else may have the information they need to fix the situation right away.
Besides CloudWatch, what are some of the other integrations you're looking to implement next?
[1] http://nagios.sourceforge.net/docs/3_0/eventhandlers.html
In a way, neptune is your virtual first-level on call engineer. If we can resolve alerts for you automatically, neptune won't bother you. If action fails for some reason, it's escalated automatically to right on call engineer.