Code: https://github.com/istio/istio/tree/master/tests/e2e/tests/mixer

Dashboards:

• Presubmit (Prow, CircleCI)
• Postsubmit (Prow, CircleCI)
• Release (Prow)

Features:

• Client-side telemetry reporting
• Ability to report telemetry for configured faults

Summary:

This test validates that (a) telemetry is generated for injected faults and (b) that the destination workload information is all unknown for that telemetry (correctly reported). This was added in response to an issue in which the mixer Report protocol implementation was not properly accounting for attribute deletion in batches. The setup and execution mirror the reproduction detailed in https://github.com/istio/istio/issues/11151.

Associated Issue(s):

• https://github.com/istio/istio/issues/11151

Associated istio.io Task(s):

None.

Test Method:

• Install Bookinfo sample
• Configure VirtualService and Destination Rule to abort 100% of calls to details service, returning http status of 555
• Send 100 requests to productpage service via ingressgateway
• Using retry logic, query Prometheus looking for at least 50 requests (istio_requests_total) that have a response flag of FI and response code of 555 with, importantly, unknown destination workload values. 50 was chosen as a reasonable bar to ensure that we are capturing telemetry about injected faults while still providing robustness in the face of slow tests, configuration delays, etc. This test is not meant to validate fault injection behavior in totality, but rather to validate that Istio does report telemetry in the face of fault injection.

Common failure modes:

No known failures in pre/post submit runs since test was added to validate fix for the reported issue.

Code: https://github.com/istio/istio/tree/master/tests/e2e/tests/dashboard

Dashboards:

• Presubmit (Prow, CircleCI)
• Postsubmit (Prow, CircleCI)
• Release (Prow)