Download as PDF, PPTX
![if [ "$1" = "deploy" ] && [ $# -eq 4 ]; then
STAGE=$2
REGION=$3
PROFILE=$4
npm install
AWS_PROFILE=$PROFILE 'node_modules/.bin/sls' deploy -s $STAGE -r $REGION
…](https://image.slidesharecdn.com/serverlessinproduction-171017152644/75/Serverless-in-production-O-Reilly-Software-Architecture-99-2048.jpg)
![if [ "$1" = "deploy" ] && [ $# -eq 4 ]; then
STAGE=$2
REGION=$3
PROFILE=$4
npm install
AWS_PROFILE=$PROFILE 'node_modules/.bin/sls' deploy -s $STAGE -r $REGION
…
install serverless framework as
dev dependency](https://image.slidesharecdn.com/serverlessinproduction-171017152644/75/Serverless-in-production-O-Reilly-Software-Architecture-100-2048.jpg)
![http://bit.ly/2k93hAj
kinesis
global.CONTEXT
log.info(…)
api-b
global.CONTEXT
global.CONTEXT
global.CONTEXT
x-correlation-id = …
x-correlation-xxx = …
API Gateway Kinesis
SNS
API Gateway
API Gatewayapi-a api-c
sns
headers[“User-Agent”]
headers[“Debug-Log-Enabled”]
MessageAttributes: [
“x-correlation-id”: …
“User-Agent”: …
“Debug-Log-Enabled”: …
]
global.CONTEXT
headers[“User-Agent”]
headers[“Debug-Log-Enabled”]
headers[“x-correlation-id”]
headers[“User-Agent”]
headers[“Debug-Log-Enabled”]
headers[“x-correlation-id”]
data.__context
capture
forward
function
event](https://image.slidesharecdn.com/serverlessinproduction-171017152644/75/Serverless-in-production-O-Reilly-Software-Architecture-126-2048.jpg)
Uploaded byYan Cui
Serverless in production (O'Reilly Software Architecture)
The document discusses the experiences and best practices for running serverless architectures in production, covering topics such as testing, CI/CD, logging, monitoring, configuration management, and managing cold starts. It emphasizes the need for small, fast, and independently deployable deployments while minimizing costs and operational effort. It also highlights various tools and approaches to enhance serverless implementation, including the importance of end-to-end testing and effective logging and monitoring strategies.
More Related Content
![Learning the Alphabet: A/B, CD and [E-Z] in the Docker Datacenter by Brett Ti...](https://cdn.slidesharecdn.com/ss_thumbnails/alphabet-160624232958-thumbnail.jpg?width=640&height=640&fit=bounds)
Similar to Serverless in production (O'Reilly Software Architecture)
![Devfest Harare 2025 Slides [GDG Harare] - Combined Slide Deck](https://cdn.slidesharecdn.com/ss_thumbnails/ilovepdfmerged12-251029161756-3b829161-thumbnail.jpg?width=640&height=640&fit=bounds)
Serverless in production (O'Reilly Software Architecture)
- 1. Serverless in production anexperience report Yan Cui
- 2. What’s in thistalk? ! how to responsibly run a serverless architecture (aka. how to do ops in serverless) ! testing, CI/CD ! logging, distributed tracing, monitoring ! config management, securing secrets ! coldstarts ! gotchas/limitations + workarounds/hacks
- 3.
- 4.
- 12.
- 13. Before ! hidden complexitiesand dependencies ! low utilisation to leave headroom for large spikes ! EC2 scaling is slow, so scale earlier ! paying for lots of used resources ! up to 30 mins to deploy ! deployments required downtime
- 14. - Dan North “leadtime to someone saying thank you is the only reputation metric that matters.”
- 16. “what would goodlook like for us?”
- 17. Deployments should… ! besmall ! be fast ! have zero downtime ! require no lock-step
- 18. Features should… ! beindependently deployable ! be loosely-coupled
- 19. We want to… !minimise cost of unused resources ! minimise ops effort ! reduce technical mess ! deliver visible improvements to users faster
- 20.
- 21. 170 Lambda functionsin prod 1.2 GB deployment packages in prod 95% cost saving vs EC2 15x no. of prod releases per month
- 22.
- 23. 1st function inprod! time is a good fit
- 24. ? time is a goodfit 1st function in prod!
- 25. Practices ToolsPrinciples what isgood? how to make it good? with what?
- 26.
- 27.
- 28. 170 functions WOOF! ? ? time isa good fit 1st function in prod!
- 29.
- 30.
- 31. building a bettersearch experience
- 32. Legacy Monolith AmazonKinesis Amazon Lambda Amazon CloudSearch
- 33. Legacy Monolith AmazonKinesis Amazon Lambda Amazon CloudSearchAmazon API Gateway Amazon Lambda
- 34. building an analyticspipeline
- 35. Legacy Monolith AmazonKinesis Amazon Lambda Google BigQuery
- 36. Legacy Monolith AmazonKinesis Amazon Lambda Google BigQuery 1 developer, 2 days design production (his 1st serverless project)
- 37. Legacy Monolith AmazonKinesis Amazon Lambda Google BigQuery “thank you, nothing ever got done this fast at Skype!”
- 38. - Dan North “leadtime to someone saying thank you is the only reputation metric that matters.”
- 39. rebuilding the timelinefeature
- 46. building better userrecommendations
- 47.
- 48.
- 49.
- 50.
- 51.
- 52.
- 53.
- 54.
- 55.
- 56.
- 57.
- 58.
- 59.
- 60.
- 61.
- 62.
- 63. Level of Testing 1.Unit doour objects do the right thing? are they easy to work with?
- 65. 1.Unit 2.Integration does our codework against code we can’t change? Level of Testing
- 66.
- 67. handler test by invokingthe handler
- 68. Level of Testing 1.Unit 2.Integration 3.Acceptance doesthe whole system work?
- 69.
- 70. “…We find thattests that mock external libraries often need to be complex to get the code into the right state for the functionality we need to exercise. The mess in such tests is telling us that the design isn’t right but, instead of fixing the problem by improving the code, we have to carry the extra complexity in both code and test…” Don’t Mock Types You Can’t Change
- 71. “…The second riskis that we have to be sure that the behaviour we stub or mock matches what the external library will actually do… Even if we get it right once, we have to make sure that the tests remain valid when we upgrade the libraries…” Don’t Mock Types You Can’t Change
- 72. Services Don’t Mock TypesYou Can’t Change
- 73. Paul Johnston The serverlessapproach to testing is different and may actually be easier. http://bit.ly/2t5viwK
- 74.
- 75.
- 76. LambdaAPI Gateway DynamoDB UnitTests Mock/Stub
- 77. is our requestcorrect? is the request mapping set up is the API resources configured correctly? are we assuming the correct schema? LambdaAPI Gateway DynamoDB is Lambda proxy configured correctly? is IAM policy set up correctly? is the table created? what unit tests will not tell you…
- 79. most Lambda functionsare simple have single purpose, the risk of shipping broken software has largely shifted to how they integrate with external services observation
- 81. But it slowsdown my feedback loop… IT’S NOT ABOUT YOU!
- 82. me test your system, not(just) your code
- 83.
- 84. …if a servicecan’t provide you with a relatively easy way to test the interface in reality, then you should consider using another one. Paul Johnston
- 85. “…Wherever possible, anacceptance test should exercise the system end-to-end without directly calling its internal code. An end-to-end test interacts with the system only from the outside: through its interface…” Testing End-to-End
- 86. Legacy Monolith AmazonKinesis Amazon Lambda Amazon CloudSearchAmazon API Gateway Amazon Lambda
- 87. Legacy Monolith AmazonKinesis Amazon Lambda Amazon CloudSearchAmazon API Gateway Amazon Lambda Test Input
- 88. Legacy Monolith AmazonKinesis Amazon Lambda Amazon CloudSearchAmazon API Gateway Amazon Lambda Test Input Validate
- 89. integration tests exercise system’sIntegration with its external dependencies
- 90. acceptance tests exercise systemEnd-to-End from the outside
- 91. integration tests differfrom acceptance tests only in HOW the Lambda functions are invoked observation
- 95.
- 96. “…We prefer tohave the end-to-end tests exercise both the system and the process by which it’s built and deployed… This sounds like a lot of effort (it is), but has to be done anyway repeatedly during the software’s lifetime…” Testing End-to-End
- 97. me Deployment scripts thatonly live on the CI box is a disaster waiting to happen.
- 98. Jenkins build configdeploys and tests unit + integration tests deploy acceptance tests
- 99. if [ "$1"= "deploy" ] && [ $# -eq 4 ]; then STAGE=$2 REGION=$3 PROFILE=$4 npm install AWS_PROFILE=$PROFILE 'node_modules/.bin/sls' deploy -s $STAGE -r $REGION …
- 100. if [ "$1"= "deploy" ] && [ $# -eq 4 ]; then STAGE=$2 REGION=$3 PROFILE=$4 npm install AWS_PROFILE=$PROFILE 'node_modules/.bin/sls' deploy -s $STAGE -r $REGION … install serverless framework as dev dependency
- 101. can be runlocally & on the CI box
- 103.
- 104.
- 106. 2016-07-12T12:24:37.571Z 994f18f9-482b-11e6-8668-53e4eab441ae GOT isoff air, what do I do now?
- 107. 2016-07-12T12:24:37.571Z 994f18f9-482b-11e6-8668-53e4eab441ae GOT isoff air, what do I do now? UTC Timestamp API Gateway Request Id your log message
- 108.
- 109. me Logs are noteasily searchable in CloudWatch Logs.
- 110.
- 111.
- 112. CENTRALISE LOGS MAKE THEMEASILY SEARCHABLE
- 113. + + the elkstack
- 114.
- 115. CloudWatch Logs AWSLambda ELK stack
- 116.
- 117.
- 118.
- 119.
- 121. “my followers didn’t receivemy new post!” - a user
- 122.
- 123. correlation IDs* * eg.request-id, user-id, yubl-id, etc.
- 124.
- 125.
- 126. http://bit.ly/2k93hAj kinesis global.CONTEXT log.info(…) api-b global.CONTEXT global.CONTEXT global.CONTEXT x-correlation-id = … x-correlation-xxx= … API Gateway Kinesis SNS API Gateway API Gatewayapi-a api-c sns headers[“User-Agent”] headers[“Debug-Log-Enabled”] MessageAttributes: [ “x-correlation-id”: … “User-Agent”: … “Debug-Log-Enabled”: … ] global.CONTEXT headers[“User-Agent”] headers[“Debug-Log-Enabled”] headers[“x-correlation-id”] headers[“User-Agent”] headers[“Debug-Log-Enabled”] headers[“x-correlation-id”] data.__context capture forward function event
- 127.
- 128.
- 129.
- 130. traces do notspan over API Gateway
- 131.
- 132. “where do Iinstall monitoring agents?”
- 133.
- 134. • invocation Count •error Count • latency • throttling • granular to the minute • support custom metrics
- 135. • invocation Count •error Count • latency • throttling • granular to the minute • support custom metrics
- 139. Why not IOPipe? !pervasive access to your entire application ! adds latency for tracking
- 140. me The only “background” processingyou get are the capabilities the platform provides out of the box.
- 141. “how do Ibatch up and send logs/metrics in the background?”
- 142.
- 143. console.log(“hydrating yubls fromdb…”); 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
- 144. CloudWatch Logs AWSLambda ELK stack logs metrics CloudWatch
- 145.
- 146. CloudWatch Logs AWSLambda ELK stack logs metrics CloudWatch memory used memory size billed duration
- 147.
- 148.
- 149.
- 150.
- 151.
- 152.
- 154. don’t put allyour eggs in one basket aka. you don’t want your monitoring system to fail at the same time as the systems it monitors
- 155.
- 156.
- 157. me Environment variables makeit hard to share configurations across functions.
- 158. me Environment variables makeit hard to implement fine-grained access to sensitive info.
- 159.
- 160. couples ability todeploy with access to sensitive data, which often don’t overlap in a large engineering team or in a regulated environment
- 161.
- 162.
- 165. Why not consulor etcd? ! multiple EC2 instances in multi-AZ for HA ! have to manage servers, patch OS, patch software, etc. ! learning curve for configuring the service ! learning curve for using the CLI tools
- 167.
- 168. SSM Parameter Store HTTPS role-basedaccess encrypted in-flight
- 169.
- 170.
- 171. HTTPS role-based access SSM ParameterStore encrypted in-flight
- 172.
- 173.
- 174. Requirements for clientlibrary ! standardise and encapsulate how you manage configs ! supports client-side caching (fetch & cache at coldstart) ! invalidate cache at interval ! invalidate cache explicitly when staleness is detected
- 175.
- 176.
- 177. max 75 GBtotal deployment package size* * limit is per AWS region
- 178.
- 179.
- 180.
- 181. install Serverless frameworkas dev dependency at project level dev dependencies are excluded since 1.16.0
- 182.
- 183.
- 184.
- 185. Amazon X-Ray 1st invocation 2ndinvocation cold start
- 186.
- 187.
- 188.
- 189.
- 190.
- 191. me C# and Javaexperiences ~100 times the cold start time of Python and also suffer from much higher standard deviation
- 192. me memory size improvescold start time linearly
- 193.
- 194. CloudWatch Event AWSLambda
- 195. CloudWatch Event AWSLambda ping ping ping ping
- 196. CloudWatch Event AWSLambda ping ping ping ping
- 197. CloudWatch Event AWSLambda ping ping ping ping HEALTH CHECKS?
- 198. AWS Lambda docs Take advantageof container re-use to improve the performance of your function. Make sure any externalized configuration or dependencies that your code retrieves are stored and referenced locally after initial execution. Limit the re-initialization of variables/objects on every invocation. Instead use static initialization/ constructor, global/static variables and singletons. Keep alive and reuse connections (HTTP, database, etc.) that were established during a previous invocation. http://amzn.to/2jzLmkb
- 199. max 5 minsexecution time
- 200.
- 201.
- 202. AWS Lambda docs AWS Lambdapolls your stream and invokes your Lambda function. Therefore, if a Lambda function fails, AWS Lambda attempts to process the erring batch of records until the time the data expires. http://amzn.to/2vs2lIg
- 203. vs processing halts untilfailed events are retried successfully/ expired from stream prioritize realtime-ness, retry failed events with best effort, then skip
- 204. SNS Kinesis SQS after 3 attempts shareprocessing logic events are processed in chronological order failed events are retried out of sequence
- 205.
- 206.
- 207.
- 208. AWS Lambda docs Each shardcan support up to 5 transactions per second for reads, up to a maximum total data read rate of 2 MB per second. http://amzn.to/2ubyaot
- 209. AWS Lambda docs If yourstream has 100 active shards, there will be 100 Lambda functions running concurrently. Then, each Lambda function processes events on a shard in the order that they arrive. http://amzn.to/2ubyaot
- 210. when no. ofprocessors goes up…
- 211. ReadProvisionedThroughputExceeded can have toomany Kinesis read operations…
- 212. ReadRecords.IteratorAge unpredictable spikes inread ‘latency’…
- 213.
- 214.
- 215.
- 216. new tool, newproblems but they’re easier to deal with
- 217.
- 218.