Lambda Invocations¶

A. `Synchronous`¶

below services make sync/blocking call:
ALB, Amazon API Gateway, CloudFront (Lambda@Edge)
Amazon S3-Batch
Amazon Cognito
AWS Step Functions
Amazon Lex
Amazon Alexa
Amazon Kinesis Data Firehose

A.1 ALB¶

Flow
http/s request comes to ELB
elb converts http request into json object and pass to event arg
call lambda-1 > handler-1(event, context)
lambda sends json object as response

ELB converts lambda-response into http-response

{
"statusCode": 200,
"statusDescription": "200 OK",
"isBase64Encoded": false,
"headers": {
  "Content-Type": "text/html",
  "Set-Cookie": "cookie1=value1; Path=/; HttpOnly",
    "Custom-Header": "value"
  },
"body": "<html><body>Hello from Lambda!</body></html>"
}

conversions:
ALB Multi-Header Values : enable/disable this feature.
notice queryparams,headers array

A.2 gateway¶

soon

---¶

B `A-synchronous`¶

below services(event-based) make a-sync/non-blocking call:
s3:evnetNotification
SQS, SNS, SES(email)
CW:log-events||subscription-filter
more (not in scope of exam)
- AWS codeCommit + codePiprline + cloudformation
- AWS config
- AWs IoT
internal eventQueue
event --> eventSQSQueue [e1,e2,...] --> LambdaService reads event and invoke lambda functions async.
if throttleError, LambdaService will return event back to internal eventQueue.

:green_circle: B.1 S3:event notification¶

create s3 bucket > properties tab >> create s3 notification
set prefix + s3:objectCreate event
set target : lambda-1 (lambda-policy-1: allow s3)
create DLQ or destination(OnFailure)

:green_circle: B.2 SQS¶

message dropped on SQS queue-1
sqs-event will call lambda async does (non-blocking)
next, lambda-1 >> configuration tab >> asynchronous invocation section.
add DLQ-1 on lambda(not on queue-1)
set retry attempt = 0,1,2
also, update lambda-role-1: + sqsSend permission
if exception thrown from lambda(consumer), then lambda will retry
keep lambda code Idempotent
after reties goes to DLQ-1

:green_circle: B.3 Event-bridge (generic pattern)¶

trigger:
EventRule
schedular

C. `Event Source Mapping` (Poller) + batch¶

udemy reference
Lambda is triggered synchronously with batch
by polling data from below 3 poll-based services
Queue based: Queue Poller : SQS : ordered, if FIFO.
streams based: Stream Poller : KDS : ordered + DynamoDB Streams

- lambda will scale out, based on active message.

:yellow_circle: C.1 SQS : Queue¶

Event Source Mapping will poll SQS (Long Polling)
configuration:
batch size + batch Window
recommended : Set the queue visibility timeout = 6x Lambda-1::timeout
set-up DLQ-1 on the SQS queue.
preferred way : destination (on-failure)
more:
For FIFO queue : lambda supports in order processing.
lambda will delete item, after processing.

:yellow_circle: C.2 KDS : stream¶

- parallelization: can have upto 10 batches per shard. - in-order processing : processing for the affected shard is paused, until the error is resolved

- shard-1
  - poll-1 > batch-1 > lambda-1-instance-1
  - poll-2 > batch-2 > **lambda-1-instance-2**   <<< errored out
  - ...
  - ...
  - poll-10 > batch-10 > lambda-1-instance-10

  if lambda-1-instance-2 excution error out, entire shard-1 will be on hold/paused   <<<

  And entire batch, batch-2 will be reprocessed.                                     <<<

- Configuration: - discard batch - will go to Destination - restrict the number of retries

:yellow_circle: C.3 DynamoDB : stream¶

- soon

D Destination (2019)¶

udemy reference
new, preferred, 4 services
sqs
sns
another-lambda,
event-bus
in ccgg
lambda > java-rest-api > programmatically writing into success fifo queue.
so achieving same behaviour.

D.1 Async invocation: `failure`¶

Destination (condition: OnFailure)
DLQ (old, only 2 service, we did before)
sqs
sns

D.2 Async invocation: `success`¶

destination (condition: OnSuccess)

Lambda Invocations¶

A. Synchronous¶