Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

ref: Introduce a SessionFlusher #279

Merged
merged 4 commits into from
Oct 16, 2020
Merged

ref: Introduce a SessionFlusher #279

merged 4 commits into from
Oct 16, 2020

Conversation

Swatinem
Copy link
Member

The flusher maintains a queue that is flushed every minute.
Sessions are batched into Envelopes at this point, but in the future
might be pre-aggregated.

The flusher maintains a queue that is flushed every minute.
Sessions are batched into Envelopes at this point, but in the future
might be pre-aggregated.
@Swatinem Swatinem requested a review from a team October 13, 2020 15:14
Copy link
Member

@jan-auer jan-auer left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This looks good, thanks!

One overall suggestion: Instead of always waiting for the flush interval, could we flush as soon as we reach the maximum item count?

Copy link
Member

@jan-auer jan-auer left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I'm realizing one more thing, although this is probably fine: Even if the application is idle or does not log any sessions, your background thread runs once every minute to check an empty queue. Since it is only once per minute, this is not that big of a deal.

The ideal solution would probably be to stop spinning if the queue is empty and wait for a CondVar which is triggered by the enqueue code. Since there's little benefit to this, I'm fine with keeping it this way.

queue.len() >= MAX_SESSION_ITEMS
};
if flush_immediately {
SessionFlusher::flush(&self.queue, &self.transport);
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This is potentially racy. Between checking the queue length and calling flush, another thread could enqueue and flush, or a flush could happen.

To avoid this, consider keeping the lock and passing a lock guard into flush.

@Swatinem Swatinem merged commit 6729dda into master Oct 16, 2020
@Swatinem Swatinem deleted the ref/session-flusher branch October 16, 2020 14:02
jan-auer added a commit that referenced this pull request Nov 25, 2020
* master: (59 commits)
  fix: Correctly apply environment from env (#293)
  fix: Make Rust 1.48 clippy happy (#294)
  docs: Document integrations and the Hub better (#291)
  ref: Remove deprecated error-chain and failure crates (#290)
  release: 0.21.0
  meta: Update Changelog
  feat: End sessions with explicit status (#289)
  fix: Scope transaction name can be overriden in sentry-actix (#287)
  fix: sentry-actix should not capture client errors (#286)
  fix: Clean up sentry-actix toml (#285)
  ref: Remove empty integrations (#283)
  feat: Add support for actix-web 3 (#282)
  feat: Preliminary work to integrate Performance Monitoring (#276)
  ref: Introduce a SessionFlusher (#279)
  fix: Set a default environment based on debug_assertions (#280)
  ref: Rearchitect the log and slog Integrations (#268)
  ref: Deprecate public fields on Integrations (#267)
  ci: Make testfast actually fast (#273)
  fix: Update surf and unbreak CI (#274)
  ci: Use smarter cache action (#272)
  ...
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

2 participants