use PhoenixProfiler on the Endpoint #58
Conversation
* main: Fix package deps Release v0.1.0 Recommend profiler on telemetry supervision tree Add a Changelog
| %Profile{data: %{metrics: metrics}} = profile | ||
|
|
||
| socket = | ||
| case profile.data[:exception] do |
There was a problem hiding this comment.
Seems like there's a race condition here, the toolbar is mounted before telemetry has a chance to collect the event [:phoenix_profiler, :endpoint, :exception] so profile.data[:exception] is always null at this point.
I'm assuming that http://localhost:4000/errors/assign-not-available should populate @exits considering
|
@leandrocp You are correct– the toolbar cannot render the error for its own request because the I think you are also correct about the race condition tho– we are racing the late write to ETS and that's why we are getting back profile data without the exception. But now I think I have an idea of how to get the tests passing, stand by! :) |
The endpoint integration tests are racing the late collection behaviour in the profiler endpoint, which sometimes results in tests failing due to receiving outdated profile information. We are attempting to correct this by fetching the profile data by its token thru another endpoint request. Hopefully this will slow things down enough to give the late write a much better chance to finish before the data is requested. In practice this isn't an actual problem, as in most cases (like the Dashboard) we poll for updated information.
|
@leandrocp Thanks for testing it out! Yes, things seem to work fine but the Heisenbugs are forcing me to question some earlier decisions. I just had a nice chat with @elbow-jason who gave me some ideas that I hope to experiment with soon! :) |
No description provided.