tailsampler: Combine batches of spans into a single batch #1864
Conversation
…nstream processors may operate on the entire trace
chris-smith-zocdoc
requested review from
bogdandrutu,
dmitryax,
james-bebbington,
nilebox,
owais,
pjanotti and
tigrannajaryan
as code owners
Codecov Report
@@ Coverage Diff @@
## master #1864 +/- ##
==========================================
+ Coverage 91.23% 91.26% +0.02%
==========================================
Files 272 272
Lines 16263 16266 +3
==========================================
+ Hits 14838 14845 +7
+ Misses 998 996 -2
+ Partials 427 425 -2
Continue to review full report at Codecov.
|
| } | ||
|
|
||
| _ = tsp.nextConsumer.ConsumeTraces(policy.ctx, allSpans) |
There was a problem hiding this comment.
Perhaps not for this PR, but in case an error happens, would be nice to have it logged, even if at debug level.
There was a problem hiding this comment.
yeah we can do this in a separate pr
processor/samplingprocessor/tailsamplingprocessor/processor_test.go
Outdated
Show resolved
Hide resolved
| } | ||
| } | ||
|
|
||
| func findTrace(a []pdata.Traces, traceID pdata.TraceID) int { |
There was a problem hiding this comment.
The name is a bit misleading, as it's not finding a trace, it's returning the number of traces that were found for the given ID. Suggestion: numTracesWithID()?
There was a problem hiding this comment.
There is probably a better way to do this, but its finding the Index of the trace in the array. I added this because there isn't a guaranteed order in which the traces are received
In the case where the trace is not found it returns the size of the array which might be confusing, I could change this to -1 if you'd like.
There was a problem hiding this comment.
Would using t.Fatal() or assert.Fail() be more appropriate since it's invalid test state?
There was a problem hiding this comment.
I've updated this method to return the trace or nil which makes it much clearer
processor/samplingprocessor/tailsamplingprocessor/processor_test.go
Outdated
Show resolved
Hide resolved
| span.SetSpanID(tracetranslator.UInt64ToByteSpanID(uint64(i + 1))) | ||
|
|
||
| spanID++ | ||
| span.SetSpanID(tracetranslator.UInt64ToByteSpanID(uint64(spanID))) |
There was a problem hiding this comment.
previously this was creating duplicate span ids, now its a monotonic sequence. the other tests don't rely on the spanid it seems
Signed-off-by: Anthony J Mirabella <a9@aneurysm9.com>
* Update core/contrib deps to v0.58.0 * Update config source provider * Add make fmt tidy and adopt fixes * create internal/tools module for project * use wget instead of now missing curl in kafka tests Co-authored-by: Ryan Fitzpatrick <rmfitzpatrick@signalfx.com>
Description:
Adding feature #1834
This changes the tail sampler to emit all spans for a single traces in a single batch. This allows downstream processors to easily perform operations on the entire trace.
Testing:
Added test