HDDS-14814. Unify Fragmented Traces for Freon randomkeys Command

[sravani-revuri]

What changes were proposed in this pull request?

Tracing for Freon randomkeys is currently fragmented and has the following issues:

• Instead of a single trace for a command execution, multiple traces are being created. Therefore breaking the parent-child hierarchy.
• The root span doesnt contain command name as it appears for ozone sh commands.
• A single freon rk command creates multiple repetitive spans.

Proposed Changes:
Proper handling of spans is required to ensure a chain of spans is created for the same trace.
Implement span level sampling through custom sampler and enable it through OTEL_SPAN_SAMPLING.

What is the link to the Apache JIRA

https://issues.apache.org/jira/browse/HDDS-14814

How was this patch tested?

Manual Testing through Jaeger UI and unit tests.

Running Regular Freon Command

Command:

ozone freon rk --numOfVolumes=1 --numOfBuckets=1 --numOfKeys=3 
OTEL_SPAN_SAMPLING=""

Output:

Running Span Sampling only

Command:

ozone freon rk --numOfVolumes=1 --numOfBuckets=4 --numOfKeys=2
OTEL_SPAN_SAMPLING=createVolume:1,createBucket:2,createKey:5

Output:

1 CreateVolume call:

2 CreateBucket calls:

1 CreateKey call:

Running Span Sampling with Trace Sampling

Command:

ozone freon rk --numOfVolumes=1 --numOfBuckets=1 --numOfKeys=10 
OTEL_TRACES_SAMPLER_ARG=0.5
OTEL_SPAN_SAMPLING=createVolume:1,createBucket:1,createKey:5

Output:

8 traces visible when command is run 16 times.

without sampling 180 spans are generated.

[Gargi-jais11]

Thanks @sravani-revuri for the patch.
Left few comments.

[sumitagrawl]

@sravani-revuri thanks for working over this, few comments given

[sumitagrawl]

LGTM

[adoroszlai]

Thanks @sravani-revuri for the patch.

[adoroszlai]

Thanks @sravani-revuri for updating the patch.

[adoroszlai]

Sorry, I just noticed this. We should increase visibility of parseSpanSamplingConfig to package-private to avoid reflection in the test.