WIP: Emit debugging events #567
Conversation
|
|
|
My employer has signed a CLA. |
Codecov Report
@@ Coverage Diff @@
## main #567 +/- ##
==========================================
+ Coverage 85.44% 85.60% +0.15%
==========================================
Files 6 6
Lines 378 382 +4
Branches 85 85
==========================================
+ Hits 323 327 +4
Misses 31 31
Partials 24 24
📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more |
|
We are now using this in production and it's been extremely helpful for us. I'd love to get this into master! |
| @@ -200,6 +200,8 @@ class SlackBot extends Adapter | |||
| # NOTE: should rate limit errors also bubble up? | |||
| if error.code isnt -1 | |||
| @robot.emit "error", error | |||
| else | |||
| @robot.emit "slack-rate-limit" | |||
There was a problem hiding this comment.
it took quite a bit of detective work, but it turns out that we don't know of any code that triggers the "error" event on the RtmClient, which is the only code path that results in this method ever getting called.
it seems like the comparison of code to -1 is remnants from the v3.x series of hubot-slack, which used the v1.x series of @slack/client, which would have emitted a "error" event.
therefore, i don't think this code is worth merging, because it would just become even more misleading.
| @@ -192,10 +192,13 @@ class SlackClient | |||
| thread_ts: envelope.message?.thread_ts | |||
|
|
|||
| if typeof message isnt "string" | |||
| @web.chat.postMessage(room, message.text, _.defaults(message, options)) | |||
| messageOptions = _.defaults(message, options) | |||
| @robot.emit "postMessage", room, message.text, messageOptions | |||
There was a problem hiding this comment.
if the goal is just to understand when chat.postMessage is called, then i don't think emitting on the robot is appropriate.
instead, i think the goal should be to turn on more logging, specifically for the WebClient. currently, there's a pattern established to set more options on the RtmClient using the HUBOT_SLACK_RTM_CLIENT_OPTS environment variable. so an alternative that i would favor is to introduce a HUBOT_SLACK_WEB_CLIENT_OPTS environment variable that can be used to set the logLevel to "debug" (and any of the other options).
what do you think?
Logging is actually not great for us! We were specifically hooking into this in order to send stats to our stats monitoring service. Parsing logs is inconvenient, but if it's emitted as an event then our user code can listen for it and emit metrics when they're received. |
|
got it! so the use case is specifically to gather data for a monitoring service. i think Hubot already provides an abstraction designed for the use case where you want to run some code before (or after) anything is sent by the adapter: response middleware. would that be sufficient for your use case? |
Summary
My company's Hubot has been facing some repeated rate-limiting issues, and we've found that hubot-slack has been missing the diagnostics we need to be able to figure out the cause. This adds some new event emitters to let user scripts listen in for these events. So far I've added emitters for every message that's posted, and whenever a rate limit occurs.
An open question for me on the latter is whether there's any other place that rate limiting might come up, and whether there's any detail that I can include in the
slack-rate-limitevent in order to allow consumers to gather meaningful, actionable data.cc @aoberoi
Requirements (place an
xin each[ ])