These are more terse than I prefer but I'm tired and want to capture some interesting points learned tonight:
- When triggering a lambda function on Put, consider enabling "Multipart upload completed" if your objects are large. If you don't see the event generated, check the size of the upload and then check this setting :)
- There are zero options when using the built-in event-to-Lambda configuration (as compared to SQS which has visibility, retry, monitoring, DLQ, etc).
- If you pipe S3 events into SQS you get the S3 events inside the body of your SQS events. It's turtles all the way down.
- Last but not least (and unrelated to S3 events): use the headObject API to fetch object metadata NOT getObject unless of course you want to burn up a lot of Lambda execution time.
- When triggering a lambda function on Put, consider enabling "Multipart upload completed" if your objects are large. If you don't see the event generated, check the size of the upload and then check this setting :)
- There are zero options when using the built-in event-to-Lambda configuration (as compared to SQS which has visibility, retry, monitoring, DLQ, etc).
- If you pipe S3 events into SQS you get the S3 events inside the body of your SQS events. It's turtles all the way down.
- Last but not least (and unrelated to S3 events): use the headObject API to fetch object metadata NOT getObject unless of course you want to burn up a lot of Lambda execution time.