Upcoming changes for LocalStack v2

Hello everyone! We are excited to announce that we are preparing a major release of LocalStack which will introduce some impactful changes, and we want to use this issue to keep you informed.

The announcement can also be found on GitHub: LocalStack v2 release 🚧 · Issue #7882 · localstack/localstack · GitHub

Tentative Timeline

  • Thursday March 23., 2023: :warning: we will merge our release branch v2 into master, which will update the latest Docker image to contain the current release candidate of v2. This will put into effect all breaking changes that we are introducing for v2, and will affect your pipeline if you are using latest.
  • Thursday March 30., 2023: Tagged release v2.0.0 will be published

What’s changed?

:construction: What follows is a draft of the release notes, which will help LocalStack users prepare for a migration ahead of time. We will update the release notes with non-critical changes until the release, and then publish final release notes as part of the release.


  • LocalStack community image (localstack/localstack) no longer contains LocalStack Pro (localstack/localstack-pro) code, effectively cutting the community image size in half
  • Removal of localstack-full and localstack-light since they have become obsolete thanks to our new dependency packaging system (simply use localstack or localstack-pro)
  • Completely new AWS Lambda provider with improved parity and performance
  • Completely new S3 provider with improved parity and features
  • Improved supports for community cloud pods
  • Simplified host configuration and docker networking
  • Internal endpoints have moved into /_localstack and /_aws


  • Cross-service IAM enforcement
  • Completely new snapshot persistence (PERSISTENCE=1) mechanism with more flexible load and save strategies
  • New simplified container setup for AWS Big Data Technologies (Athena, Glue, EMR, etc)

How to migrate



  • We are unifying the variables EDGE_BIND_HOST, EDGE_PORT and EDGE_PORT_HTTP into GATEWAY_LISTEN, which will allow configuration of the addresses and ports the LocalStack process listens on. It takes the form <ip address>[:port][,<ip address>:<port>...] where multiple values can be given, separated by commas. LocalStack will listen on all interfaces, but will ask for superuser permission for privileged ports. We will still accept EDGE_BIND_HOST, EDGE_PORT, and EDGE_PORT_HTTP during a deprecation period following the release of v2, but they will not be used by LocalStack for configuration at some point in the future. Please migrate your use of EDGE_BIND_HOST, EDGE_PORT or EDGE_PORT_HTTP to GATEWAY_LISTEN. If you experience issues, please try the deprecated configuration variables.

  • For example, if you previously ran LocalStack with the command EDGE_BIND_HOST= EDGE_PORT=5000 EDGE_PORT_HTTP=5001 localstack start or used these configuration variables in your docker-compose.yml, please replace it with GATEWAY_LISTEN=0.0.0:5000, localstack start.

  • We are unifying the variables HOSTNAME_EXTERNAL and LOCALSTACK_HOSTNAME into LOCALSTACK_HOST, which will allow configuration of hostnames returned by LocalStack in a more consistent way. If provided, this variable will be used in preference to HOSTNAME_EXTERNAL and LOCALSTACK_HOSTNAME, and will be used as the hostname returned in URLs for created resources such as OpenSearch clusters, SQS queues, or RDS databases. We will still accept HOSTNAME_EXTERNAL or LOCALSTACK_HOSTAME during a deprecation period following the release of v2, but they will not be used by LocalStack for configuration at some point in the future. Please migrate your use of HOSTNAME_EXTERNAL or LOCALSTACK_HOSTNAME to LOCALSTACK_HOST. If you experience issues, please try the deprecated configuration variables.

  • For example, if you previously ran LocalStack with the command HOSTNAME_EXTERNAL=<hostname> localstack start or used HOSTNAME_EXTERNAL in your docker-compose.yml, please replace it with LOCALSTACK_HOST=<hostname>[:port].

  • Starting LocalStack Pro using the CLI will no longer publish port 53 to the host if it is already bound by some other service like systemd-resolved on Linux, or mDNSResponder on macOS.

Persistence & cloud pods

Using LocalStack Pro with PERSISTENCE=1 (which we term snapshot-based persistence) now has two behavioral changes with respect to when data is restored and saved. More information can be found in our docs.

  • Restoring state/loading a snapshot (load): Previously, snapshots could only be loaded from disk per-service when services were first initialized, lazy-loading the state the first time a service was used. With the new persistence mechanism introduced in v2, persistent data can be loaded on LocalStack startup. You can configure this behavior by setting SNAPSHOT_LOAD_STRATEGY to on_request or on_startup. The default strategy currently is still on_request.
  • Saving state/creating a snapshot (save): Previously, a snapshot was created for a particular service on each request. This was mainly to protect against potential data loss if LocalStack would suddenly terminate. We found that this approach leads to several problems, specifically related to concurrency and performance. An alternative approach we have introduced is to store snapshots on LocalStack shutdown, which produces no performance overhead during runtime, but will not protect you against data loss if LocalStack does not terminate correctly. The default strategy is on a scheduled basis, specifically, we take snapshots of services that have changed every 15 seconds. You can configure this behavior by setting SNAPSHOT_SAVE_STRATEGY to on_request, on_shutdown or scheduled, respectively.

The layout of Cloud Pods has been changed and cloud pods created with v1.x.x may be incompatible with v2.0. Please re-create your cloud pods with the latest LocalStack version.


Lambda has been completely re-written and the current documentation can be found here. There are several behavioral changes to lambda that mostly affect users of the LOCAL executor mode. If you run into issues in v2, you can, for now, use the legacy provider using PROVIDER_OVERRIDE_LAMBDA=legacy.

  • Mounting the Docker socket "/var/run/docker.sock:/var/run/docker.sock" into the LocalStack container is now required to run Lambdas.
  • The default behavior is now equivalent to the old docker-reuse executor, there are no longer multiple options.
  • Local executor: with LAMBDA_EXECUTOR=local , lambdas were executed within the LocalStack container. This was mostly used as a fallback if the docker socket could not be mounted into the LocalStack container. If you cannot mount the docker socket or don’t have an external DOCKER_HOST available, we provide a new way to run Lambda functions via static worker containers that can be configured manually. The only requirement here is connectivity between the static worker and the LocalStack instance. Predefined workers will be available via images hosted by us: e.g. localstack/lambda-worker:python3.9.
  • There are several new ways to configure the lambda provider variables that can be found in our docs, which we will update in the upcoming weeks.
  • We have migrated our Lambda Docker images from lambci to use the official AWS images, which will now be pulled by default from public.ecr.aws/lambda/
  • Functions were previously created synchronously, i.e., CreateFunction calls would block until the function state was “Active”. Functions are now created asynchronously, and their state will move from “Pending” to “Active”, which you can check with a GetFunction call. This behavior can be disabled with LAMBDA_SYNCHRONOUS_CREATE=1 (not recommended).
  • Stricter input validation: previously, when creating functions, the Role attribute could be any value, and many of our examples included something like awslocal create-function --role r1. This will no longer work, as roles now have to have an ARN format. We do not t validate whether the role exists, so you can use any ARN.
  • Hot reloading: previously, the magic S3 bucket name for lambda code hot reloading was __local__, which was changed to hot-reload. Please change your deployment configs accordingly, or use the BUCKET_MARKER_LOCAL configuration to customize the value.
  • Lambda in LocalStack Pro supports “transparent endpoint injection”, which allows Lambdas to resolve domains like s3.amazonaws.com to the LocalStack container. Previously, this was based on patching SDKs, but is now completely DNS-based, and will be disabled if DNS_ADDRESS=0 is set.


S3 has been completely re-written and its behavior aligned with AWS. Users should be mostly unaffected, but may experience some breakage if depending on previous behavior that was not aligned with AWS. Should you run into problems in v2, you can activate the old provider with PROVIDER_OVERRIDE_S3=legacy.

Big Data container

  • The mono container mode (BIGDATA_MONO_CONTAINER=1) for BigData services (Glue, EMR, Athena, etc) is now the default, the previous implementation with a separate sidecar localstack_bigdata container is deprecated and will be removed in a future release.
  • Please note that some of the required dependencies (Spark, Hive, etc) are lazily downloaded and installed at runtime, which increases the processing time on first load. The libraries are cached locally in the var_libs directory - please make sure to properly mount the LocalStack volume into your container. We also provide a separate localstack/localstack-pro:latest-bigdata BigData mono container image which has the default dependencies pre-installed.

Other notable changes

  • supervisord is no longer used to start LocalStack in the Docker container, instead we use our own init program.
  • LocalStack no longer automatically restarts on failure
  • /var/lib/localstack/logs/localstack_infra.log was removed
  • REQUIRE_PRO has been replaced with ACTIVATE_PRO, which sets whether or not LocalStack pro should be activated when using the pro image.

Deprecation removals

  • Deprecation removals:
    • LEGACY_EDGE_PROXY has been removed
    • LEGACY_DIRECTORIES has been removed, please migrate to the current Filesystem Layout
    • LEGACY_IAM_ENFORCEMENT has been removed
    • /docker-entrypoint-initaws.d for initializing LocalStack has been removed, please migrate to the modern Initialization Hooks
    • please find other minor deprecation notices in previous release notes.