Currently if the ingestor function take a long time to run (due to an error in the pipeline) there is no limit on the number of concurrent executions, mean the function can scale up and cost a lot.
Configuring a reserved concurrency limit in the CDK configuration should solve this. Currently the function hovers around 30-40 concurrent executions so limiting to 100 would provide a safe ceiling.
Currently if the ingestor function take a long time to run (due to an error in the pipeline) there is no limit on the number of concurrent executions, mean the function can scale up and cost a lot.
Configuring a reserved concurrency limit in the CDK configuration should solve this. Currently the function hovers around 30-40 concurrent executions so limiting to 100 would provide a safe ceiling.