Closed rf closed 8 years ago
@rf Let's do straight sampling for v0.
We want to verify that sampling makes error frames faster with our benchmarks.
I'm working on https://github.com/uber/tchannel/pull/1262
I think we can keep the ordering semantics per worker somehow. But if we lose it is not a problem. Let's just get the ts
correct on the LogRecord so we can re-order if needed.
Lets get some tests in here too :)
Let's add stats then ship.
Fix lint.
Let's also get a stat for everything we flush
In production I want to see the distribution of logs that are dropped and flushed by levelName.
legit. please fix travis though.
@Raynos @jcorbin @kriskowal
To continue the discussion on the commit:
record.js
. It uses an object with the data that doesn't change and copies the meta fields into it. My benchmarks showed that to be faster, but I could remove the optimization for clarity and instead just copy all fields into a new object.