Containerd sensor releases

This release improves log collection when handling container shutdown scenarios.

Previously, lookup for additional labels present in container images on Kubernetes failed when CRI-API was used. This issue is now fixed. The lookup now uses imageRef as the image identifier.

The Containerd sensor now supports the native CRI-API client that you can enable with the environment variable INSTANA_USE_CONTAINERD_CRI_CLIENT (default: true only in Kubernetes). When set to true, the metrics are collected by the cgroup and the ContainerD socket that is typically located at /var/run/containerd/containerd.sock.

This release fixes an issue that is related to retrieval of Total Cache and Total RSS through the agent service for both cgroupv1 and cgroupv2.

This release provides the following fixes:

• ContainerD log sensor instances now shut down as expected to prevent duplicate log collection when monitored containers restart within Kubernetes pods or deployments.
• Error handling from client-side sensor misconfiguration is now improved.
• Dropping of some logs during the container startup and teardown process is now resolved.

The Containerd sensor now supports custom poll rate configuration.

You can now parse the cgroup data directly by using the environment variable INSTANA_CONTAINERD_PARSE_CGROUP_DATA=true .

The Containerd sensor now supports Total Normalized CPU Time and Memory Working Set Usage metrics for Control Group V2 (cgroupv2).

The Containerd sensor now supports the following Containerd versions: 1.6.x, 1.7.x, and 2.0.x.

This release adds support for heartbeat to the Containerd sensor, which reduces the data that is sent to the Instana backend. The agent only sends data every 10 seconds if the backend supports the heartbeat.