[processor/probabilistic_sampler] The default sampling mode in documentation is self-contradictory

[namco1992]

Component(s)

processor/probabilisticsampler

Describe the issue you're reporting

In the probabilistic_sampler README, it mentions that the hash_seed is selected by default:

opentelemetry-collector-contrib/processor/probabilisticsamplerprocessor/README.md

Line 146 in 964a652

This mode is selected by default.

However, later in the configuration sector, it says

the default is "proportional" unless either hash_seed is configured or attribute_source is set to record.

opentelemetry-collector-contrib/processor/probabilisticsamplerprocessor/README.md

Line 307 in 964a652

- `mode` (string, optional): One of "proportional", "equalizing", or "hash_seed"; the default is "proportional" unless either `hash_seed` is configured or `attribute_source` is set to `record`.

According to the code below, the default mode is indeed hash_seed. The const defaultMode is also set to hash_seed, which makes the following logic always select hash_seed, regardless of whether the HashSeed value is set or not:

opentelemetry-collector-contrib/processor/probabilisticsamplerprocessor/sampler_mode.go

Lines 388 to 397 in 964a652

	if mode == modeUnset {
	// Reasons to choose the legacy behavior include:
	// (a) having set the hash seed
	// (b) logs signal w/o trace ID source
	if cfg.HashSeed != 0 || (isLogs && cfg.AttributeSource != traceIDAttributeSource) {
	mode = HashSeed
	} else {
	mode = defaultMode
	}
	}

According to the #31894, it mentioned:

The default sampling mode remains HashSeed. We consider a future change of default to Proportional to be desirable.

I suppose we need a follow-up to decide if we want to switch the default sampling mode and correct the document accordingly to avoid any confusion. I'd love to open a PR for it, but I'm unsure how we want to proceed.

[github-actions]

Pinging code owners:

• processor/probabilisticsampler: @jpkrohling @jmacd

See Adding Labels via Comments if you do not have permissions to add labels yourself.

[jmacd]

I think we should correct the documentation. Thanks @namco1992 for the investigation.

[namco1992]

Got it, will PR it in, thanks!