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

Experimental: Concurrently flush records in goroutines #28

Closed
wants to merge 29 commits into from

Conversation

PettitWesley
Copy link
Contributor

Issue #, if available:
Related to: fluent/fluent-bit#2159

Description of changes:
Previously, this plugin was synchronous. Since Fluent Bit is single-threaded, this decreases performance. Native C plugins in Fluent Bit core can take advantage of its event loop to asynchronously make requests. Go plugins are not as fancy. With this change, the plugin returns control to Fluent Bit immediately and sends data in a goroutine. I've previously verified that the goroutines are proper separate threads and continue running even after control returns to the Fluent Bit C code.

The downside to this change is that plugins will not accurately report errors or retries to Fluent Bit. I'm considering making this change an optional mode that is turned off by default and users can opt into if they want.

Long term of course, we will rewrite this plugin C and contribute it to the core of Fluent Bit, giving us performance and concurrency without the above drawback.

By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of your choice.

@sonofachamp
Copy link
Contributor

The downside to this change is that plugins will not accurately report errors or retries to Fluent Bit.

Is there a correctness tradeoff as well in that Fluent Bit will not be able to retry sending the data beyond what our plugin is capable of? It seems we're just returning an FLB_OK when retries are exhausted.

PettitWesley and others added 4 commits May 22, 2020 00:36
Fixes duplicate logs when kinesis limits are exceeded.

Reduces the likelihood of dropped logs when kinesis provides backpressure (improvements in retries).

Adds log messages to indicate when and how many records are dropped due to retry timeouts.

Signed-off-by: Zack Wine <[email protected]>
@PettitWesley PettitWesley changed the base branch from master to mainline June 19, 2020 16:27
@PettitWesley
Copy link
Contributor Author

Closing in favor of #33

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.

3 participants