Pitchfork.fork_sibling: handle the middle process getting stuck #74
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
That middle process only forks again and then exit, as such it should be very fast. However because of after fork callbacks (on `Process._fork`) and `at_exit` callbacks, it's possible that a bug cause it to get stuck or be very slow. To be resilient to this kind of bug, we hard kill the middle process if it didn't exit in 2 seconds. And to make it easier to debug, we kill it with `SIGBUS` so it produce a crash report with a backtrace, core dump etc.
DazWorrall
approved these changes
Oct 31, 2023
casperisfine
pushed a commit
to casperisfine/opentelemetry-ruby
that referenced
this pull request
Oct 31, 2023
Ref: open-telemetry#363 Ref: Shopify/pitchfork#74 At Shopify we're using the Pitchfork web server, and one thing it does is that it sometimes do a "double fork", similar to the classic daemonization dance. Recently we noticed that sometimes the "middle process" of that daemonization dance would get "stuck". After adding some debug we figured it's because we have: ```ruby at_exit do # Triggers flushing of buffer on shutdown OpenTelemetry.tracer_provider.shutdown(timeout: 60) end ``` So the child process inherits lots of spans from its parents, and normally OpenTelemetry deal gracefully with that by droping the inherited spans. But here because we called `shutdown`, `@keep_running` is set to `false` hence `reset_on_fork` is not called. I don't understand why we shouldn't call it during shutdown though, aside from avoiding to respawn the thread. With this change we expect our "middle process" to drop all the inherited spans and to have nothing to flush.
fbogsany
added a commit
to open-telemetry/opentelemetry-ruby
that referenced
this pull request
Oct 31, 2023
#1537) * BatchSpanProcessor#force_flush: purge inherited spans even on shutdown Ref: #363 Ref: Shopify/pitchfork#74 At Shopify we're using the Pitchfork web server, and one thing it does is that it sometimes do a "double fork", similar to the classic daemonization dance. Recently we noticed that sometimes the "middle process" of that daemonization dance would get "stuck". After adding some debug we figured it's because we have: ```ruby at_exit do # Triggers flushing of buffer on shutdown OpenTelemetry.tracer_provider.shutdown(timeout: 60) end ``` So the child process inherits lots of spans from its parents, and normally OpenTelemetry deal gracefully with that by droping the inherited spans. But here because we called `shutdown`, `@keep_running` is set to `false` hence `reset_on_fork` is not called. I don't understand why we shouldn't call it during shutdown though, aside from avoiding to respawn the thread. With this change we expect our "middle process" to drop all the inherited spans and to have nothing to flush. * Appease the cop --------- Co-authored-by: Jean Boussier <[email protected]> Co-authored-by: Francis Bogsanyi <[email protected]>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
That middle process only forks again and then exit, as such it should be very fast.
However because of after fork callbacks (on
Process._fork) andat_exitcallbacks, it's possible that a bug cause it to get stuck or be very slow.To be resilient to this kind of bug, we hard kill the middle process if it didn't exit in 2 seconds.
And to make it easier to debug, we kill it with
SIGBUSso it produce a crash report with a backtrace, core dump etc.