Skip to content

Latest commit

 

History

History
281 lines (201 loc) · 15.3 KB

CHANGELOG.md

File metadata and controls

281 lines (201 loc) · 15.3 KB

Changelog

Deprecations

History

Unreleased

New

  • Can use Pod Identity with Azure Event Hub scaler (#994)
  • Introducing InfluxDB scaler (#1239)
  • Add Redis cluster support for Redis list and Redis streams scalers (#1437)
  • Global authentication credentials can be managed using ClusterTriggerAuthentication objects (#1452)
  • Introducing OpenStack Swift scaler (#1342)

Improvements

  • Support add ScaledJob's label to its job (#1311)
  • Bug fix in aws_iam_authorization to utilize correct secret from env key name (PR #1332)
  • Add metricName field to postgres scaler and auto generate if not defined. (PR #1381)
  • Mask password in postgres scaler auto generated metricName. (PR #1381)
  • Bug fix for pending jobs in ScaledJob's accurateScalingStrategy . (#1323)
  • Fix memory leak because of unclosed scalers. (#1413)
  • Override the vhost on a RabbitMQ scaler via vhostName in the metadata. (#1451)
  • Optimize Kafka scaler's getLagForPartition function. (#1464)
  • Reduce unnecessary /scale requests from ScaledObject controller (#1453)
  • Add support for the WATCH_NAMESPACE environment variable to the operator (#1474)
  • Automatically determine the RabbitMQ protocol when possible, and support setting the protocl via TriggerAuthentication (#1459,#1483)
  • Improve performance when fetching pod information (#1457)
  • Improve performance when fetching current scaling information on Deployments (#1458)

Breaking Changes

Other

  • Bump go module version to v2 (#1324)

v2.0.0

New

  • KEDA uses a dedicated HTTP client, connection pool, and (optional) TLS certificate for each configured scaler
  • KEDA scales any CustomResource that implements Scale subresource (#703)
  • Provide KEDA go-client (#494)
  • Define KEDA readiness and liveness probes (#788)
  • KEDA Support for configurable scaling behavior in HPA v2beta2 (#802)
  • Add External Push scaler (#820 | docs)
  • Managed Identity support for Azure Monitor scaler (#936)
  • Add support for multiple triggers on ScaledObject (#476)
  • Add consumer offset reset policy option to Kafka scaler (#925)
  • Add option to restore to original replica count after ScaledObject's deletion (#219)
  • Add Prometheus metrics for KEDA Metrics API Server (#823 | docs)
  • Add support for multiple redis list types in redis list scaler (#1006) | docs)
  • Introduce Azure Log Analytics scaler (#1061) | docs)
  • Add Metrics API Scaler (#1026)
  • Add cpu/memory Scaler (#1215)
  • Add Scaling Strategy for ScaledJob (#1227)
  • Add IBM MQ Scaler (#1253)

Improvements

  • Move from autoscaling v2beta1 to v2beta2 for HPA (#721)
  • Introduce shortnames for CRDs (#774)
  • Improve kubectl get scaledobject to show related trigger authentication (#777)
  • Improve kubectl get triggerauthentication to show information about configured parameters (#778)
  • Added ScaledObject Status Conditions to display status of scaling (#750)
  • Added optional authentication parameters for the Redis Scaler (#962)
  • Improved GCP PubSub Scaler performance by closing the client correctly (#1087)
  • Added support for Trigger Authentication for GCP PubSub scaler (#1291)

Breaking Changes

  • Change apiGroup from keda.k8s.io to keda.sh (#552)
  • Introduce a separate ScaledObject and ScaledJob(#653)
  • Remove New() and Close() from the interface of service ExternalScaler in externalscaler.proto.
  • Removed deprecated brokerList for Kafka scaler (#882)
  • All scalers metadata that is resolved from the scaleTarget environment have suffix FromEnv added. e.g: connection -> connectionFromEnv
  • Kafka: split metadata and config for SASL and TLS (#1074)
  • Service Bus: queueLength is now called messageCount (#1109)
  • Use host instead of apiHost in rabbitmq scaler. Add protocol in trigger spec to specify which protocol should be used (#1115)
  • CRDs are using apiextensions.k8s.io/v1 apiVersion (#1202)

Other

  • Change API optional structs to pointers to conform with k8s guide (#1170)
  • Update Operator SDK and k8s deps (#1007,#870,#1180)
  • Change Metrics Server image name from keda-metrics-adapter to keda-metrics-apiserver (#1105)
  • Add Argo Rollouts e2e test (#1234)

v1.5.0

Learn more about our release in our milestone.

New

  • Scalers
    • Introduce Active MQ Artemis scaler (Docs)
    • Introduce Redis Streams scaler (Docs | Details)
    • Introduce Cron scaler (Docs | Details)
  • Secret Providers
    • Introduce HashiCorp Vault secret provider (Docs | Details)
  • Other
    • Introduction of nodeSelector in raw YAML deployment specifications (Details)

Improvements

  • Improved message count determination when using includeUnacked in RabbitMQ scaler (Details)
  • Fix for blank path without trailing slash in RabbitMQ scaler (Details)
  • Improved parsing of connection strings to support BlobEndpoint, QueueEndpoint, TableEndpoint & FileEndpoint segments (Details)
  • Support scaling when no storage checkpoint exists in Azure Event Hubs scaler (Details)
  • GCP Pub Scaler should not panic on invalid credentials (Details)
  • Make queueLength optional in RabbitMQ scaler (Details)

Breaking Changes

None.

Other

None.

v1.4.1

New

None

Improvements

  • Fix for scale-to-zero for Prometheus scaler no longer working (#770)
  • Fix for passing default VHost for Rabbit MQ scaler no longer working (#770)
  • Provide capability to define time encoding for operator (#766)

Breaking Changes

None.

Other

  • Print version of metric adapter in logs (#770)

v1.4.0

New

  • Extend RabbitMQ scaler to support count unacked messages(#700)

Improvements

  • Fix scalers leaking (#684)
  • Provide installation YAML package as release artifact (#740)
  • Improve Azure Monitor scaler to handle queries without metrics (#680)
  • Authenticate to AWS with dedicated role without AssumeRole permissions (#656)
  • KEDA now respects label restrictions on Horizontal Pod Autoscaler to have max 63 chars (#707)
  • KEDA will automatically assign deploymentName label if it was not defined in ScaledObject (#709)

Breaking Changes

None.

Other

  • Adding label for metrics service selection (#745)
  • Filter returned metrics from api server based on queried name (#732)
  • Add redis host and port parameter to the scaler with tests (#719)
  • Remove go micro version (#718)
  • Update zero result return to be non-error inducing (#695)
  • Return if kafka offset response is nil (#689)
  • Fix typos in MySQL scaler (#683)
  • Update README to mention CNCF (#682)

v1.3.0

New

  • Add Azure monitor scaler (#584)
  • Introduce changelog (#664)
  • Introduce support for AWS pod identity (#499)

Improvements

  • Make targetQueryValue configurable in postgreSQL scaler (#643)
  • Removed the need for deploymentName label (#644)
  • Adding Kubernetes recommended labels to resources (#596)

Breaking Changes

None.

Other

  • Updating license to Apache per CNCF donation (#661)

v1.2.0

New

  • Introduce new Postgres scaler (#553)
  • Introduce new MySQL scaler (#564)
  • Provide SASL_SSL Plain authentication for Kafka trigger scalar to work with Event Hubs (#585)

Improvements

  • TLS parameter to Redis-scaler (#540)
  • Redis db index option (#577)
  • Optional param for ConfigMaps and Secrets (#562)
  • Remove manually adding sslmode to connection string (#558)
  • ScaledObject.Status update should handle stale resource (#582)
  • Improve reconcile loop (#581)
  • Address naming changes for postgresql scaler (#593)

Breaking Changes

None.

Other

  • Move Metrics adapter into the separate Deployment (#506)
  • Fix gopls location (#574)
  • Add instructions on local development and debugging (#583)
  • Add a checkenv target (#600)
  • Mentioning problem with checksum mismatch error (#605)

v1.1.0

New

  • Introduce new Huawei Cloud CloudEye scaler (#478)
  • Introduce new kinesis stream scaler (#526)
  • Introduce new Azure blob scaler (#514)
  • Support for SASL authentication for Kafka scaler (#486)

Improvements

  • Event Hub scalar expansion to work with Java and C# applications (#517)
  • Escape Prometheus querystring (#521)
  • Change how number of pending messages is calculated and add more error handling. (#533)
  • Service bus scaler pod identity fix (#534)
  • Eventhub scalar fix (#537)
  • Kafka scaler fix for SASL plaintext auth (#544)

Breaking Changes

None.

Other

  • ScaledObject Status clean up (#466)
  • Add default log level for operator (#468)
  • Ensure get the metrics that have been aggregated (#509)
  • Scale from zero when minReplicaCount is > 0 (#524)
  • Total running Jobs must not exceed maxScale - Running jobs (#528)
  • Check deploymentName definition in ScaledObject (#532)

v1.0.0

New

  • Many more scalers added
  • Scaler extensibility (run scalers in a different container and communicate with KEDA via gRPC)
  • TriggerAuthentication and Pod Identity for identity based auth that can be shared across deployments
  • Schedule jobs on events in addition scaling out deployments

Improvements

  • Additional tests and automation through GitHub Actions

Breaking Changes

  • RabbitMQ host property now must resolve from a secret (#347)

Other

None.