1. how to build
Serverless
OBSERVABILITY
into a application

3. Abraham Wald

4. Abraham Wald

5. Abraham Wald

6. Abraham Wald
Wald noted that the study only
considered the aircraft that had survived
their missions—the bombers that had
been shot down were not present for the
damage assessment.
The holes in the returning aircraft, then,
represented areas where a bomber could
take damage and still return home safely.

7. Abraham Wald
Wald noted that the study only
considered the aircraft that had survived
their missions—the bombers that had
been shot down were not present for the
damage assessment.
The holes in the returning aircraft, then,
represented areas where a bomber could
take damage and still return home safely.

8. survivor bias in monitoring

9. survivor bias in monitoring
Only focus on failure modes that we were able to successfully
identify through investigation and postmortem in the past.
The bullet holes that shot us down and we couldn’t identify stay
invisible, and will continue to shoot us down.

10. What do I mean by “observability”?

11. Monitoring
watching out for
known failure modes
in the system,
e.g. network I/O, CPU,
memory usage, …

12. Observability
being able to debug
the system, and gain
insights into the
system’s behaviour

13. In control theory, observability is a measure of how well
internal states of a system can be inferred from
knowledge of its external outputs.
https://en.wikipedia.org/wiki/Observability

14. Known Success

15. Known SuccessKnown Errors

16. Known SuccessKnown Errors
easy to monitor!

17. Known SuccessKnown Errors
Known Unknowns

18. Known SuccessKnown Errors
Known UnknownsUnknown Unknowns

19. Known SuccessKnown Errors
Known UnknownsUnknown Unknowns
invisible bullet
holes

20. Known SuccessKnown Errors
Known UnknownsUnknown Unknowns

21. Known SuccessKnown Errors
Known UnknownsUnknown Unknowns
alert on this

22. Known SuccessKnown Errors
Known UnknownsUnknown Unknowns
alert on the
absence of this!

23. Known SuccessKnown Errors
Known UnknownsUnknown Unknowns
what went wrong?

24. These are the four pillars of the Observability Engineering
team’s charter:
• Monitoring
• Alerting/Visualization
• Distributed systems tracing infrastructure
• Log aggregation/analytics
“
” http://bit.ly/2DnjyuW- Observability Engineering at Twitter

25. microservices death stars circa 2015

26. microservices death stars circa 2015
mm… I wonder what’s
going on here…

31. microservices death stars circa 2015
I got this!

34. hi, my name is Yan.

35. I’m a principal engineer at

39. available in:
Austria, Switzerland, Germany,
Japan, Canada and Italy
30+ platforms

40. over 500,000 concurrent viewers

41. coming to the US

43. We’re hiring! Visit
engineering.dazn.com
to learn more.
follow @dazneng for
updates about the
engineering team

44. follow @dazneng for
updates about the
engineering team
We’re hiring! Visit
engineering.dazn.com
to learn more.
WE’RE HIRING!

46. AWS user since 2009

47. http://bit.ly/yubl-serverless

48. http://bit.ly/production-ready-serverless

49. since July 2018

51. new
challenges

52. NO ACCESS
to underlying OS

53. NOWHERE
to install agents/daemons

54. •nowhere to install agents/daemons
new challenges

55. user request
user request
user request
user request
user request
user request
user request
critical paths:
minimise user-facing latency
handler
handler
handler
handler
handler
handler
handler

56. user request
user request
user request
user request
user request
user request
user request
critical paths:
minimise user-facing latency
StatsD
handler
handler
handler
handler
handler
handler
handler
rsyslog
background processing:
batched, asynchronous, low
overhead

57. user request
user request
user request
user request
user request
user request
user request
critical paths:
minimise user-facing latency
StatsD
handler
handler
handler
handler
handler
handler
handler
rsyslog
background processing:
batched, asynchronous, low
overhead
NO background processing
except what platform provides

58. •no background processing
•nowhere to install agents/daemons
new challenges

59. EC2
concurrency used to be
handled by your code

60. EC2
Lambda
Lambda
Lambda
Lambda
Lambda
now, it’s handled by the
AWS Lambda platform

61. EC2
logs & metrics used to be
batched here

62. EC2
Lambda
Lambda
Lambda
Lambda
Lambda
now, they are batched in each
concurrent execution, at best…

65. HIGHER concurrency to log
aggregation/telemetry system

66. •higher concurrency to telemetry system
•nowhere to install agents/daemons
•no background processing
new challenges

67. Lambda
cold start

68. Lambda
data is batched between
invocations

69. Lambda
idle
data is batched between
invocations

70. Lambda
idle
garbage collectiondata is batched between
invocations

71. Lambda
idle
garbage collectiondata is batched between
invocations
HIGH chance of data loss

72. •high chance of data loss (if batching)
•nowhere to install agents/daemons
•no background processing
•higher concurrency to telemetry system
new challenges

73. Lambda

74. my code
send metrics

75. my code
send metrics

76. my code
send metrics
internet internet
press button something happens

78. http://bit.ly/2Dpidje

79. ?
functions are often chained together
via asynchronous invocations

80. ?
SNS
Kinesis
CloudWatch
Events
CloudWatch
LogsIoT
DynamoDB
S3 SES

81. ?
SNS
Kinesis
CloudWatch
Events
CloudWatch
LogsIoT
DynamoDB
S3 SES
tracing ASYNCHRONOUS
invocations through so many
different event sources is difficult

82. •asynchronous invocations
•nowhere to install agents/daemons
•no background processing
•higher concurrency to telemetry system
•high chance of data loss (if batching)
new challenges

83. These are the four pillars of the Observability Engineering
team’s charter:
• Monitoring
• Alerting/Visualization
• Distributed systems tracing infrastructure
• Log aggregation/analytics
“
” http://bit.ly/2DnjyuW- Observability Engineering at Twitter

84. LOGGING

86. 2016-07-12T12:24:37.571Z 994f18f9-482b-11e6-8668-53e4eab441ae
GOT is off air, what do I do now?

87. 2016-07-12T12:24:37.571Z 994f18f9-482b-11e6-8668-53e4eab441ae
GOT is off air, what do I do now?
UTC Timestamp Request Id
your log message

89. one log group per
function
one log stream for each
concurrent invocation

90. logs are not easily searchable in
CloudWatch Logs
me

91. CloudWatch Logs

93. CloudWatch Logs is an async event source for Lambda

94. Concurrent Executions
Time
regional max
concurrency
functions that are
delivering business value

95. Concurrent Executions
Time
regional max
concurrency
functions that are
delivering business value
ship logs

96. either set concurrency limit on the log shipping function
(and potentially lose logs due to throttling)
or…

98. 1 shard = 1 concurrent execution
i.e. control the no. of concurrent
executions with no. of shards

99. …

100. CloudWatch Logs

106. CloudWatch Logs

107. use structured logging with JSON

108. https://stackify.com/what-is-structured-logging-and-why-developers-need-it/ https://blog.treasuredata.com/blog/2012/04/26/log-everything-as-json/

109. https://www.loggly.com/blog/8-handy-tips-consider-logging-json/

111. traditional loggers are too heavy for Lambda

115. CloudWatch Logs
$0.50 per GB ingested
$0.03 per GB archived per month

116. CloudWatch Logs
$0.50 per GB ingested
$0.03 per GB archived per month
1M invocation of a 128MB function =
$0.000000208 * 1M + $0.20 =
$0.408

117. DON’T leave debug logging ON in production

119. have to redeploy ALL the
functions along the call path to
collect all relevant debug logs

124. EC2
Lambda
Lambda
Lambda
Lambda
Lambda
Concurrency is handled by
the AWS Lambda platform

128. sampling decision has to be
followed by an entire call chain

129. Initial Request ID
User ID
Session ID
User-Agent
Order ID
…

130. nonintrusive
extensible
consistent
works for streams

131. EC2
Lambda
Lambda
Lambda
Lambda
Lambda
Concurrency is handled by
the AWS Lambda platform

132. store correlation IDs in global variable

135. use middleware to auto-capture incoming correlation IDs

136. extract correlation IDs from
invocation event, and store them in
the correlation-ids module
reset

138. logger to always include captured correlation IDs

139. HTTP and AWS SDK clients to auto-forward correlation IDs on

140. context.awsRequestId
get-index

141. context.awsRequestId x-correlation-id
get-index

142. {
“headers”: {
“x-correlation-id”: “…”
},
…
}
get-index

143. {
“body”: null,
“resource”: “/restaurants”,
“headers”: {
“x-correlation-id”: “…”
},
…
}
get-index get-restaurants

144. get-restaurants
global.CONTEXT
global.CONTEXT
x-correlation-id = …
x-correlation-xxx = …
get-index
headers[“User-Agent”]
headers[“Debug-Log-Enabled”]
headers[“User-Agent”]
headers[“Debug-Log-Enabled”]
headers[“x-correlation-id”]
capture
forward
function
event
log.info(…)

147. nonintrusive
extensible
consistent
works for streams

148. MONITORING

150. •no background processing
•nowhere to install agents/daemons
new challenges

151. my code
send metrics
internet internet
press button something happens

152. those extra 10-20ms for
sending custom metrics would
compound when you have
microservices and multiple
APIs are called within one slice
of user event

153. Amazon found every 100ms of latency cost them 1% in sales.
http://bit.ly/2EXPfbA

154. console.log(“hydrating yubls from db…”);
console.log(“fetching user info from user-api”);
console.log(“MONITORING|1489795335|27.4|latency|user-api-latency”);
console.log(“MONITORING|1489795335|8|count|yubls-served”);
timestamp metric value
metric type
metric namemetrics
logs

155. CloudWatch Logs AWS Lambda
ELK stack
logs
m
etrics
CloudWatch

157. delay
cost
concurrency

158. delay
cost
concurrency
no latency
overhead

159. API Gateway
send custom metrics
asynchronously

160. SNS KinesisS3API Gateway
…
send custom metrics
asynchronously
send custom metrics as
part of function invocation

163. TRACING

164. X-Ray

168. don’t span over async
invocations
good for identifying dependencies of a function,
but not good enough for tracing the entire call
chain as user request/data flows through the
system via async event sources.

169. don’t span over non-AWS services

173. write structured logs

174. instrument your code

175. make it easy to do the right thing

176. Yan Cui
http://theburningmonk.com
@theburningmonk

177. follow @dazneng for
updates about the
engineering team
We’re hiring! Visit
engineering.dazn.com
to learn more.
WE’RE HIRING!