knative/eventing v0.18.3

Knative Eventing release v0.18.3

on GitHub

Action Required

• You must run pre-install job prior to upgrading to get apiserversource to v1beta1 API. (#4031, @capri-xiyue)
• Change storage version of remaining messaging.- resources from v1beta1 to v1:
  • “Subscriptions.messaging.knative.dev”
  • You must run pre-install job prior to upgrading to get messaging.- resources to v1 API. (#3936, @aliok)
• Change storage versions of pingsource resource from v1alpha2 to v1beta1 API
  • You must run pre-install job prior to upgrading to get pingsource resources to v1beta2 API. (#3951, @aavarghese)
• Change storage versions of eventing.- resources from v1beta1 to v1:
  • “brokers.eventing.knative.dev”
  • “Triggers.eventing.knative.dev”
  • You must run pre-install job prior to upgrading to get eventing.- resources to v1 API. (#3923, @aliok)
• Change storage versions of flows.- resources from v1beta1 to v1:
  • “Channels.messaging.knative.dev”
  • “Inmemorychannels.messaging.knative.dev”
  • You must run pre-install job prior to upgrading to get messaging.- resources to v1 API. (#3925, @aliok)
• Change storage versions of flows.- resources from v1beta1 to v1:
  • “Parallels.flows.knative.dev”
  • “Sequences.flows.knative.dev”
  • You must run pre-install job prior to upgrading to get flows.- resources to v1 API. (#3924, @aliok)

New Features

• Allow MTChannelBroker TTL to be configured via ENV variable. (#3962, @vaikas)
• PingSource adapter now uses bucket-based leader election (#4009, @lionelvillard)
• PingSource adapter deployment can now be customized at installation time (#3987, @lionelvillard)

Bug Fixes

• The default exponential backoff duration now matches the advertised algorithm in the DeliverySpec.BackoffDelay comments. User action required to evaluate if DeliverySpec.BackoffDelay settings in Subscription.spec.delivery remain appropriate. (#4115, @travis-minke-sap)
• Channels and Brokers correctly retry on non 2xx HTTP status codes (#4112, @pierDipi)
• Make pingsource adapter controller read-only (#4099, @lionelvillard)
• If you create a Parallel, then later add branches to it, caused a panic. (#3946, @vaikas)
• In cases where Filter sends a message and it fails or response is nil, it will panic because it uses it. (#3897, @vaikas)
• Tests sometimes flake when the webhook fails. (#3906, @vaikas)
• The subscription reconciler correctly propagates delivery configurations to the channel (retry, backoffDelay, backoffPolicy). (#4042, @pierDipi)
• Trigger reconciler correctly reconciles Triggers during parallel updates of the Trigger and the referenced Broker. (#3966, @pierDipi)
• When not reconciling triggers not owned by my brokerclass, would print incorrect msg, logic was correct however, just wrong log message. (#4030, @vaikas)
• Update the spec to include v1 of the channelable. (#3870, @vaikas)
• Fix bug which could cause flakiness in the IMC because IMC Dispatcher was configured for normal reconciler instead of readonly. This could cause races. (3686, @vaikas). (4280, @vaikas)
• Fix bug which could cause Trigger Dependency flakiness because we were incorrectly setting up the informer tracking the dependent sources. (4294, @vaikas). (4296, @vaikas)