Slide 1

Slide 1 text

Blrdroid Meetup, Nov 2024 Mobile Observability with OpenTelemetry Jitin Sharma, Groww

Slide 2

Slide 2 text

Observability

Slide 3

Slide 3 text

System Inputs Outputs

Slide 4

Slide 4 text

Sum Machine 2,2 4

Slide 5

Slide 5 text

Sum Machine 2,2 4 4,4 8

Slide 6

Slide 6 text

Sum Machine 2,2 4 4,4 6,4 8 -1

Slide 7

Slide 7 text

Sum Machine 2,2 4 4,4 6,4 8 -1

Slide 8

Slide 8 text

Mobile Apps

Slide 9

Slide 9 text

No content

Slide 10

Slide 10 text

Mobile App User Action UI

Slide 11

Slide 11 text

Mobile App Backend

Slide 12

Slide 12 text

Mobile App Cell Tower CDN/ Gateway Backend

Slide 13

Slide 13 text

Mobile App Cell Tower CDN/ Gateway Backend • OEM • Performance • OS Versions • Network state • User lifecycle

Slide 14

Slide 14 text

Mobile App Cell Tower CDN/ Gateway Backend • OEM • Performance • OS Versions • Network state • User lifecycle Background Jobs Noti fi cations

Slide 15

Slide 15 text

Mobile App Cell Tower CDN/ Gateway Backend • OEM • Performance • OS Versions • Network state • User lifecycle Background Jobs Noti fi cations Backend Code

Slide 16

Slide 16 text

What could go wrong

Slide 17

Slide 17 text

Issue due to wrong code

Slide 18

Slide 18 text

Issue due to wrong code Network fl uctuations

Slide 19

Slide 19 text

Issue due to wrong code Network fl uctuations Lifecycle actions

Slide 20

Slide 20 text

Issue due to wrong code Network fl uctuations Lifecycle actions Device

Slide 21

Slide 21 text

Issue due to wrong code Network fl uctuations Lifecycle actions Device OEM

Slide 22

Slide 22 text

Issue due to wrong code Network fl uctuations Lifecycle actions Device OEM Pyramid of problems

Slide 23

Slide 23 text

Issue due to wrong code Network fl uctuations Lifecycle actions Device OEM Pyramid of problems

Slide 24

Slide 24 text

Existing tools

Slide 25

Slide 25 text

System Inputs Outputs

Slide 26

Slide 26 text

System Inputs Outputs

Slide 27

Slide 27 text

Now what

Slide 28

Slide 28 text

Start End P1 P2 P3 P4

Slide 29

Slide 29 text

Custom Events App Monitoring - Crashes - Performance - … App Events - Navigation - Attributes - Custom events - …

Slide 30

Slide 30 text

Problems? • Disconnected systems • Sampling • Delays and interval con fi guration • Data privacy • Costs

Slide 31

Slide 31 text

OpenTelemetry

Slide 32

Slide 32 text

What is OpenTelemetry - Framework to collect telemetry data - Vendor agnostic - Works across frameworks using agent libraries - Allows control over export of data

Slide 33

Slide 33 text

OpenTelemetry Traces Metrics Logs

Slide 34

Slide 34 text

Span Start End t0 tN attributes

Slide 35

Slide 35 text

Span - Network Call Request Start • user identi fi er • os version • device model • network type Response Received

Slide 36

Slide 36 text

Trace Span Span Span Span Span Span Span

Slide 37

Slide 37 text

Trace Span Span Span Span Span Span Span trace id

Slide 38

Slide 38 text

Distributed Tracing System 1 System 2 System 3 Spans Spans Spans Trace

Slide 39

Slide 39 text

Distributed tracing on Mobile?

Slide 40

Slide 40 text

Mobile App UI

Slide 41

Slide 41 text

Mobile App Network Storage UI System Device

Slide 42

Slide 42 text

Mobile App Network Storage UI System Retro fi t Workmanager Sqlite Shared Pref Recyclerview Device Lifecycle Memory CPU

Slide 43

Slide 43 text

Mobile App Network Storage UI System Retro fi t Workmanager Sqlite Shared Pref Recyclerview Device Lifecycle Memory CPU Server Gateway Kafka DB

Slide 44

Slide 44 text

Mobile App Network Storage UI System Retro fi t Workmanager Sqlite Shared Pref Recyclerview Device Lifecycle Memory CPU Server Gateway Kafka DB Span Span Span Span Span Span

Slide 45

Slide 45 text

No content

Slide 46

Slide 46 text

No content

Slide 47

Slide 47 text

Additional Data Metrics - Counter - Timer - Histogram - Gauge Logs - Bridge to existing log libraries

Slide 48

Slide 48 text

Show me code

Slide 49

Slide 49 text

// Build instance val openTelemetry = otelRumBuilder.build().openTelemetry // Create a tracer openTelemetry.getTracer("network.call") // Create span val span = tracer.spanBuilder(spanName) .setAllAttributes(attributes) .startSpan()

Slide 50

Slide 50 text

// Create span processor val processor = BatchSpanProcessor.builder( OtlpHttpSpanExporter.builder() .setEndpoint(endPoint) .build() ) // Attach processor to instance otelRumBuilder.addTracerProviderCustomizer { builder -> builder.addSpanProcessor(processor) }

Slide 51

Slide 51 text

// Build instance val openTelemetry = otelRumBuilder.build().openTelemetry // Create meter val meter = openTelemetry.getMeter("success.responses") // Build counter metric meter.counterBuilder(name) .setDescription(description) .setUnit(unit) .build()

Slide 52

Slide 52 text

val reader = PeriodicMetricReader.builder( OtlpHttpMetricExporter.builder() .setEndpoint(endPoint) .build() ) otelRumBuilder.addMeterProviderCustomizer { builder -> builder.registerMetricReader(reader) }

Slide 53

Slide 53 text

OpenTelemetry Android

Slide 54

Slide 54 text

Auto Instrumentation • App Launch trace • Lifecycle traces • Network call traces • Slow rendering traces • Crash detection • Network change detection • Disk persistence • Automatic device attributes

Slide 55

Slide 55 text

Data Visualisation

Slide 56

Slide 56 text

Visualisations (OTLP)

Slide 57

Slide 57 text

Visualisations (OTLP) Otel Collector

Slide 58

Slide 58 text

Why OpenTelemetry

Slide 59

Slide 59 text

• Powerful API • Open and mature ecosystem • Total control over data • Collaborative development • Observability beyond crashes

Slide 60

Slide 60 text

Is it stable?

Slide 61

Slide 61 text

Is it stable?

Slide 62

Slide 62 text

Do you need advanced observability?

Slide 63

Slide 63 text

• It depends • Without scale, existing solutions might just work. • Ecosystem under active development

Slide 64

Slide 64 text

https://github.com/open-telemetry/opentelemetry-android

Slide 65

Slide 65 text

Fin