Knative Eventing release v0.5.0
Pre-release
Pre-release
·
1 commit
to release-0.5
since this release
| Artifact | Description |
|---|---|
eventing.yaml |
Just the core knative/eventing components without Sources or ChannelProvisioners |
gcp-pubsub.yaml |
Just the GCP PubSub ChannelProvisioner |
in-memory-channel.yaml |
Just the in-memory ChannelProvisioner |
kafka.yaml |
Just the Kafka ChannelProvisioner |
natss.yaml |
Just the NATS Streaming ChannelProvisioner |
release.yaml |
Core components bundled with the in-memory ChannelProvisioner |
Eventing
Action Required
- Optional: manually delete the unused
eventing-controller-adminClusterRoleBinding. #986 - Any existing Triggers with
Anyvalues will need to be updated to the empty string instead. #985
New Features
- Introduce Broker and Trigger as event delivery mechanisms. See docs here and here for more info. #788
- A new channel-default-controller will monitor every channel. It will check if ClusterChannelProvisioner is installed and that the channel-provisioner-specific controller monitors the channel. If provisioner is not installed or corrupt and no controller monitors this channel, then the ProvisionerInstalled condition in status will be set to False and the channel will be marked as not ready. If the provisoner is installed and monitoring the channel then the ProvisionerInstalled condition in status will be set to True. #843
- Kafka Channels can configure their replication factor with the
ReplicationFactorargument. #869 - Added
consumer_modeconfig option in config mapkafka-channel-controller-config, default value ismultiplexwhich retains existing behavior, if you setconsumer_modeto bepartitions, it creates a go channel per partition for the consumer when dispatching the events to your service. #879 - Add DEFAULT_NATSS_URL and DEFAULT_CLUSTER_ID configuration to NATS provisioner #881
- Introduce a new aggregated ClusterRole for Addressable. #1013
- CRDs are now labelled with
knative.dev/crd-install=trueto allow installing without race conditions or kubectl errors. #1007
Bug Fixes
- NATSS channels now recover from connection loss. #796
- Changed old behavior for kafka channel that existing kafka consumers are stopped and resubscribed when fanout configs are updated, now kafka consumers are reused until corresponding subscription gets deleted.
#880 - Add 1GiB memory limit to controller and webhook deployments #921
- Send Trigger responses back into Broker. #917
- Support tracing through Broker and Trigger, including replies. #936 #949
- Expose metrics from the Broker ingress via a Prometheus endpoint. #937
- An Event can flow through a Broker via Trigger replies only 255 times before getting dropped. #951 #1016
- RBAC: Webhook now runs as
eventing-webhookservice account. Controller now runs aseventing-controllerservice account. Controller and webhook run with less privileges (was usingcluster-admin) #872 - RBAC: knative-eventing-webhook does not have create rule anymore for eventing.knative.dev resources. #987
- Switch from using subscriber.dnsName to subscriber.uri. subscriber.dnsName will be removed in the next release. #994