Appsignal is very very good! I highly recommend it for Elixir projects if you only use elixir and need something setup fast. It's awesome!

I also highly recommend looking into spans and tracing (if you use multiple languages together), then you can use something like spandex with elixir and other languages to do tracing across languages. It's excellent with datadog.

Looks very cool. I've been using Sentry, but this is something that may make me switch to AppSignal at some point.

It's so good to see so many improvements/cool features in Elixir/Erlang ecosystem.

If interested in setting up CI/CD for Elixir, this might be worth checking:

https://news.ycombinator.com/item?id=22236049

WombatOAM seems similar:

https://www.erlang-solutions.com/products/wombatoam.html

I've been very happy with AppSignal for my Elixir projects (introduced them to 3 paying customers).

I especially like their request-based pricing model. It makes it a no brainer to put it on every rails/elixir project under one account (per paying customer).

Disclaimer: they have sent me a box of delicious cookies for creating an open source plugin (ember deploy plugin for automatically uploading source maps to AppSignal). But I have liked them way before that :)

We're using Appsignal and it's been great. Fantastic support too.

Looks like a web version of Erlangs "observer" app, offhand.

It's not, observer is still different.

AppSignal's monitoring shows useful metrics but does not allow examining the supervisor tree and other things possible with Observer and other OTP tooling. In some sense, it's not a live debugging tool but a monitoring dashboard.

There's a web based observer, it's called wobserver https://github.com/shinyscorpion/wobserver

Yes, but it's woefully out of date.

Fun fact -- Ian (maintainer) is/was a backend dev at Square Enix.

The guy behind the Cowboy web server was working on one a while back and it looked super cool, but I don't know what became of it.

I tried out AppSignal for Elixir and was disappointed that the BEAM metrics don’t include scheduler utilization, which is kind of a proxy for CPU usage. IIRC there was scheduler count but not utilization metrics.

this is a little too much handwaving for my tastes.