This repository has been archived by the owner on Nov 7, 2022. It is now read-only.
-
Notifications
You must be signed in to change notification settings - Fork 60
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
* Add trace head-sampling This is to complete the sampling feature since OC Service already offers tail-sampling. This is implemented using an extra hashing as an attempt to avoid bias from trace ID generation and also to correctly spread traces for backends that also do hashing. Tests to enforce hash of different lengths. Make hashing seed configurable * Direct recommendation about hash seeding setting * More precise wording for hash-seed comment
- Loading branch information
Paulo Janotti
authored
Jun 19, 2019
1 parent
610d5f6
commit dee9655
Showing
6 changed files
with
559 additions
and
1 deletion.
There are no files selected for viewing
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
168 changes: 168 additions & 0 deletions
168
processor/tracesamplerprocessor/tracesamplerprocessor.go
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,168 @@ | ||
| // Copyright 2019, OpenCensus Authors | ||
| // | ||
| // Licensed under the Apache License, Version 2.0 (the "License"); | ||
| // you may not use this file except in compliance with the License. | ||
| // You may obtain a copy of the License at | ||
| // | ||
| // http://www.apache.org/licenses/LICENSE-2.0 | ||
| // | ||
| // Unless required by applicable law or agreed to in writing, software | ||
| // distributed under the License is distributed on an "AS IS" BASIS, | ||
| // WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. | ||
| // See the License for the specific language governing permissions and | ||
| // limitations under the License. | ||
|
|
||
| package tracesamplerprocessor | ||
|
|
||
| import ( | ||
| "context" | ||
| "errors" | ||
| "fmt" | ||
|
|
||
| tracepb "github.com/census-instrumentation/opencensus-proto/gen-go/trace/v1" | ||
| "github.com/spf13/viper" | ||
|
|
||
| "github.com/census-instrumentation/opencensus-service/consumer" | ||
| "github.com/census-instrumentation/opencensus-service/data" | ||
| "github.com/census-instrumentation/opencensus-service/processor" | ||
| ) | ||
|
|
||
| const ( | ||
| // The constants below are tags used to read the configuration via viper. | ||
| samplingPercentageCfgTag = "sampling-percentage" | ||
| hashSeedCfgTag = "hash-seed" | ||
|
|
||
| // The constants help translate user friendly percentages to numbers direct used in sampling. | ||
| numHashBuckets = 0x4000 // Using a power of 2 to avoid division. | ||
| bitMaskHashBuckets = numHashBuckets - 1 | ||
| percentageScaleFactor = numHashBuckets / 100.0 | ||
| ) | ||
|
|
||
| // TraceSamplerCfg has the configuration guiding the trace sampler processor. | ||
| type TraceSamplerCfg struct { | ||
| // SamplingPercentage is the percentage rate at which traces are going to be sampled. Defaults to zero, i.e.: no sample. | ||
| // Values greater or equal 100 are treated as "sample all traces". | ||
| SamplingPercentage float32 | ||
| // HashSeed allows one to configure the hashing seed. This is important in scenarios where multiple layers of collectors | ||
| // have different sampling rates: if they use the same seed all passing one layer may pass the other even if they have | ||
| // different sampling rates, configuring different seeds avoids that. | ||
| HashSeed uint32 | ||
| } | ||
|
|
||
| // InitFromViper updates TraceSamplerCfg according to the viper configuration. | ||
| func (tsc *TraceSamplerCfg) InitFromViper(v *viper.Viper) (*TraceSamplerCfg, error) { | ||
| if v == nil { | ||
| return nil, errors.New("v is nil") | ||
| } | ||
| if err := v.UnmarshalKey(samplingPercentageCfgTag, &tsc.SamplingPercentage); err != nil { | ||
| return nil, fmt.Errorf("failed to unmarshal %q: %v", samplingPercentageCfgTag, err) | ||
| } | ||
| if err := v.UnmarshalKey(hashSeedCfgTag, &tsc.HashSeed); err != nil { | ||
| return nil, fmt.Errorf("failed to unmarshal %q: %v", hashSeedCfgTag, err) | ||
| } | ||
| return tsc, nil | ||
| } | ||
|
|
||
| type tracesamplerprocessor struct { | ||
| nextConsumer consumer.TraceConsumer | ||
| scaledSamplingRate uint32 | ||
| hashSeed uint32 | ||
| } | ||
|
|
||
| var _ processor.TraceProcessor = (*tracesamplerprocessor)(nil) | ||
|
|
||
| // NewTraceProcessor returns a processor.TraceProcessor that will perform head sampling according to the given | ||
| // configuration. | ||
| func NewTraceProcessor(nextConsumer consumer.TraceConsumer, cfg TraceSamplerCfg) (processor.TraceProcessor, error) { | ||
| if nextConsumer == nil { | ||
| return nil, errors.New("nextConsumer is nil") | ||
| } | ||
|
|
||
| return &tracesamplerprocessor{ | ||
| nextConsumer: nextConsumer, | ||
| // Adjust sampling percentage on private so recalculations are avoided. | ||
| scaledSamplingRate: uint32(cfg.SamplingPercentage * percentageScaleFactor), | ||
| hashSeed: cfg.HashSeed, | ||
| }, nil | ||
| } | ||
|
|
||
| func (tsp *tracesamplerprocessor) ConsumeTraceData(ctx context.Context, td data.TraceData) error { | ||
| scaledSamplingRate := tsp.scaledSamplingRate | ||
| if scaledSamplingRate >= numHashBuckets { | ||
| return tsp.nextConsumer.ConsumeTraceData(ctx, td) | ||
| } | ||
|
|
||
| sampledTraceData := data.TraceData{ | ||
| Node: td.Node, | ||
| Resource: td.Resource, | ||
| SourceFormat: td.SourceFormat, | ||
| } | ||
|
|
||
| sampledSpans := make([]*tracepb.Span, 0, len(td.Spans)) | ||
| for _, span := range td.Spans { | ||
| // If one assumes random trace ids hashing may seems avoidable, however, traces can be coming from sources | ||
| // with various different criterias to generate trace id and perhaps were already sampled without hashing. | ||
| // Hashing here prevents bias due to such systems. | ||
| if hash(span.TraceId, tsp.hashSeed)&bitMaskHashBuckets < scaledSamplingRate { | ||
| sampledSpans = append(sampledSpans, span) | ||
| } | ||
| } | ||
|
|
||
| sampledTraceData.Spans = sampledSpans | ||
|
|
||
| return tsp.nextConsumer.ConsumeTraceData(ctx, sampledTraceData) | ||
| } | ||
|
|
||
| // hash is a murmur3 hash function, see http://en.wikipedia.org/wiki/MurmurHash. | ||
| func hash(key []byte, seed uint32) (hash uint32) { | ||
| const ( | ||
| c1 = 0xcc9e2d51 | ||
| c2 = 0x1b873593 | ||
| c3 = 0x85ebca6b | ||
| c4 = 0xc2b2ae35 | ||
| r1 = 15 | ||
| r2 = 13 | ||
| m = 5 | ||
| n = 0xe6546b64 | ||
| ) | ||
|
|
||
| hash = seed | ||
| iByte := 0 | ||
| for ; iByte+4 <= len(key); iByte += 4 { | ||
| k := uint32(key[iByte]) | uint32(key[iByte+1])<<8 | uint32(key[iByte+2])<<16 | uint32(key[iByte+3])<<24 | ||
| k *= c1 | ||
| k = (k << r1) | (k >> (32 - r1)) | ||
| k *= c2 | ||
| hash ^= k | ||
| hash = (hash << r2) | (hash >> (32 - r2)) | ||
| hash = hash*m + n | ||
| } | ||
|
|
||
| // TraceId and SpanId have lengths that are multiple of 4 so the code below is never expected to | ||
| // be hit when sampling traces. However, it is preserved here to keep it as a correct murmur3 implementation. | ||
| // This is enforced via tests. | ||
| var remainingBytes uint32 | ||
| switch len(key) - iByte { | ||
| case 3: | ||
| remainingBytes += uint32(key[iByte+2]) << 16 | ||
| fallthrough | ||
| case 2: | ||
| remainingBytes += uint32(key[iByte+1]) << 8 | ||
| fallthrough | ||
| case 1: | ||
| remainingBytes += uint32(key[iByte]) | ||
| remainingBytes *= c1 | ||
| remainingBytes = (remainingBytes << r1) | (remainingBytes >> (32 - r1)) | ||
| remainingBytes = remainingBytes * c2 | ||
| hash ^= remainingBytes | ||
| } | ||
|
|
||
| hash ^= uint32(len(key)) | ||
| hash ^= hash >> 16 | ||
| hash *= c3 | ||
| hash ^= hash >> 13 | ||
| hash *= c4 | ||
| hash ^= hash >> 16 | ||
|
|
||
| return | ||
| } |
Oops, something went wrong.