Dapr 1.15.10

This update includes bug fixes:

• Sidecar Panic when calling endpoints with redirect behavior

Sidecar Panic when calling endpoints with redirect behavior

Problem

The Dapr sidecar would crash (panic) when invoking a non-Dapr HTTP endpoint that returned a redirect response.

Impact

This bug caused the Dapr sidecar to become unstable and fail unexpectedly when interacting with external services that use HTTP redirects.

Root cause

The panic was caused by a nil pointer dereference in the HTTP metrics middleware, which occurred when an uninitialized response writer was accessed during a redirect.

Solution

The pathMatchingRW struct was updated to correctly handle redirects and always provide a valid http.Header map, which prevents the nil pointer dereference.

This is the release candidate 1.15.10-rc.1

This is the release candidate 1.16.0-rc.3

Dapr 1.15.9

This update includes bug fixes:

• Ignore 404 errors in transactional Cosmos DB delete operations

Ignore 404 errors in transactional Cosmos DB delete operations

Problem

During transactional operations, Dapr sometimes handles deletes on keys that don't exist in Cosmos DB. This is expected behavior, but it was returning errors unnecessarily.

Impact

Transactional operations could fail when attempting to delete non-existent keys, even though this is a valid scenario that should not cause errors.

Root cause

The Cosmos DB state store component was not properly handling 404 errors during transactional delete operations, treating them as actual errors instead of expected behavior.

Solution

Modified the Cosmos DB state store to ignore 404 errors during transactional delete operations, unless there's an explicit requirement for etag checking.

This is the release candidate 1.16.0-rc.2

This is the release candidate 1.15.9-rc.1

Dapr 1.15.8

This update includes bug fixes:

• Prevent Jobs from not triggering upserting in high throughput scenarios

Prevent Jobs from not triggering upserting in high throughput scenarios

Problem

Upserting a Job (or Actor reminder) in high job trigger throughput scenarios could result in the Job not being triggered.

Impact

Jobs (or Actor reminders) could not be triggered in high throughput scenarios.

Root cause

Scheduler would find a conflict in jobs being triggered between the old and new job state, and would result in no more jobs to be triggered on that host.

Solution

Ignore the conflict in the scheduler when upserting a job (or Actor reminder) in high throughput scenarios.

This is the release candidate 1.15.8-rc.1

This is the release candidate 1.16.0-rc.1

Dapr 1.15.7

This update includes bug fixes:

• Fix Workflow Start Time
• Fix use of DynamoDB as Workflow state store

Fix Workflow Start Time

Problem

Using the Workflow "Start Time" feature would result in the Workflow starting immediately.

Impact

A workflow could not be scheduled to start at some point in the future.

Root cause

The "Start Time" was not being propagated to the Actor reminder responsible for executing a Workflow instance.

Solution

Correctly propagate the Start Time of a Workflow to the Actor reminder responsible for executing a Workflow instance.

Fix use of DynamoDB as Workflow state store

Problem

Using DynamoDB as the Workflow state store would result in an error when trying to run a Workflow.

Impact

It would not be possible to run Workflows using DynamoDB as the state store.

Root cause

The state store implementation used String (S) attributes rather than the Binary (B) attributes resulting in the state not being readable.

Solution

Updated the DynamoDB state store implementation to use Binary (B) attributes for storing Workflow state.