My production app is all-of-the-sudden full of logs like those shown below. I also opened a support ticket if that's helpful.

Apr 26 11:15:45 app/web.1 agentmon: flush: send: upstream service replied with status=401: "dynoidmap: no dyno information exists for that UUID\n"
Apr 26 11:15:53 app/worker.1 agentmon: flush: send: upstream service replied with status=401: "dynoidmap: no dyno information exists for that UUID\n"

One way to solve this is to download the agent on dyno boot instead.
This closes the window to upgrade to 24h max due to dyno cycling behavior.
But does leave us to the mercy of future changes in that cycling behavior.
And there is no real way to roll out a change immediately (security issue maybe?), should that be necessary (beyond scripting up finding all apps with the feature flag and restarting all of their dynos).

Another option is to have agentmon be self updating, where it polls a know location for some meta-data about the current version and compares it to it's version, then downloads and update and exists (or execs the update).

Other thoughts?

Maybe we shouldn't do this?

$ heroku run bash -a jvm-metrics-sample-app
Running bash on ⬢ jvm-metrics-sample-app... up, run.3025 (Standard-1X)
  % Total    % Received % Xferd  Average Speed   Time    Time     Time  Current
                                 Dload  Upload   Total   Spent    Left  Speed
100   609    0   609    0     0   9536      0 --:--:-- --:--:-- --:--:--  9666
100 2210k  100 2210k    0     0  10.8M      0 --:--:-- --:--:-- --:--:-- 10.8M
agentmon setup took 0 seconds
agentmon: Listening on :14996...
agentmon: debug: handling incoming statsd packet
~ $

I've been trying to make sense of this buildpack and I'm stuck on this line:

https://github.com/heroku/heroku-buildpack-metrics/blob/master/.profile.d/heroku-metrics-daemon.sh#L32

AGENTMON_FLAGS=("-statsd-addr=:${PORT}")

So, we're telling agentmon to boot up and scrape/forward stats from statsd, and statsd is hosted on port $PORT? But isn't this the same $PORT env var that the web app is going to bind to?

• Internal ticket: https://help.heroku.com/790553
• Followup with an explanation on how to get it working: https://heroku.support/850138

When we include this buildpack on a ruby project then the application is deployed without a web dyno (and thus doesn't function)

If we remove the buildpack and redeploy everything returns to normal.

We had a weird bug this week where our application was down and we couldn't figure out why.

Using run:inside to get a console we saw a ton of agentmon failures. Once I figured out that it might be coming from the metrics buildpack I went ahead and removed it. The app was available and working after another release.

/cc @heroku/tools @daneharrigan

Hi!

The README says:

This buildpack does two things.

1. Copies a jar into the slug
2. Copies a .profile.d/ script into your slug

However after #32 it no longer does (1), only (2).

• Provision a bucket: https://github.com/heroku/foundation/issues/2313
• Update publishing scripts to publish to s3
• Update buildpack to use s3 bucket