this blog post by @decimalator is a keen window into what it's like to be a determined user in a confusing forest of tools, improving your observability step by step, testing vendor marketing against your own experience (and trying not to go broke in the meantime). 🐝❤️
i love this for a few extra big reasons. i love that it's by a self-identified ops engineer, who's in there instrumenting the shit out of their rails apps. ✊
i love that they're using triggers to do sophisticated callbacks for incident response. I love this bubbleup snap.
i love that they're using triggers to do sophisticated callbacks for incident response. I love this bubbleup snap.
And I love this most of all -- that they bootstrapped off our free tier and solved real, hard problems without spending a ton of money.
Honeycomb has quietly become the most cost-effective option, as other vendors have been larding it up with acquisitions and multiple products.
Honeycomb has quietly become the most cost-effective option, as other vendors have been larding it up with acquisitions and multiple products.
You shouldn't *need* multiple products to achieve observability. You should be able to go from very high level trends down to slicing and dicing of raw events, or view by trace, all within one simple tool.
Every time you add a tool, you are paying to store that data AGAIN.
Every time you add a tool, you are paying to store that data AGAIN.
Worse, it interrupts the flow of your debugging. If you have to copy paste an id from one tool to the next, or if you have to shift mental models from "how metrics work" to "searching through a log" to "show me a trace" -- that's cognitively expensive.
So most people just won't do it.
It's not *reasonable* to expect engineers to learn how to be proficient in four separate tools that are needed to understand the lifecycle of the request, or different tools for the system in aggregate vs the request in particular.
It's not *reasonable* to expect engineers to learn how to be proficient in four separate tools that are needed to understand the lifecycle of the request, or different tools for the system in aggregate vs the request in particular.
The trend in monitoring tools is definitely towards complixifying the shit out of their pricing model and trying to extract every last penny out of their users. It would make me *furious* as a user. Sigh
