Use a DynamoDB Stream to compute dependencies in the background and store call counts in hourly buckets.
The assumption is that span data arrives mostly ordered and a relatively large batch window and size will include mostly complete traces. If a trace isn't in this specific batch, it was already write and can be fetched. To validate this completely missing fetches and hit rate inside a batch will be exposed as metrics.
Missing
[x] Basic tests for the lambda
[x] Create a lambda zip as release artefact
[x] Update terraform to deploy the lambda and necessary role changes
Use a DynamoDB Stream to compute dependencies in the background and store call counts in hourly buckets.
The assumption is that span data arrives mostly ordered and a relatively large batch window and size will include mostly complete traces. If a trace isn't in this specific batch, it was already write and can be fetched. To validate this completely missing fetches and hit rate inside a batch will be exposed as metrics.
Missing