Docker-Provider

collector 03/16/2021

Error handling for invalid selectors

Uses k8s label and field packages for parsing and matching selectors.

Official bits for CPU performance improvements for Container Insights.

*Dev bits

• Scale tested

07/15/2020 -

Version microsoft/oms:ciprod07152020 Version mcr.microsoft.com/azuremonitor/containerinsights/ciprod:ciprod07152020 (linux)

Version microsoft/oms:win-ciprod05262020-2 Version mcr.microsoft.com/azuremonitor/containerinsights/ciprod:win-ciprod05262020-2 (windows)

Code change log

• Following hotfixes which are applicable only for Linux agent
  • Fix the issue related to collection of multi-containers in pod for the ContainerInventory table
  • Fix the containerhostname field value to have podname rather than nodename in ContainerInventory table
  • Fix OOM issue during container startup if there are high number of pods or containers on the node
  • Fix the ContainerName field value same as before in ContainerInventory table
• We didn't rebuild windows container, so the image version for windows container stays the same as last release (ciprod:win-ciprod05262020-2) before this hotfix

Windows Daemonset - Collection of Windows std/stderr logs
More Alerable Metrics (going to Metrics Store/custom metrics - see Customer Impact section below for metrics list)
Fix OOM-ing at high prometheus scrape volume
Update fluentbit (0.14.4 to 1.4.2)
Drop non-numeric metrics thru Telegraf
Reduce Health exception (when API server response is nil)
Add 'Computer' dimension to all telemetry (internal use)
Support for specifiying HTTP & HTTPS Proxy for outbound/egress (applicable only for non-AKS clusters)
Move to rbac.authorization.k8s.io/v1 for ClusterRole & ClusterRoleBinding
Reverted version change from apiextensions.k8s.io/v1 to apiextensions.k8s.io/v1beta1 for Health CRD since this require schema update
Dont ingest per container telemetry exceptions related to container inventory since these will be huge

Windows Daemonset - Collection of Windows std/stderr logs
More Alerable Metrics (going to Metrics Store/custom metrics - see Customer Impact section below for metrics list)
Fix OOM-ing at high prometheus scrape volume
Update fluentbit (0.14.4 to 1.4.2)
Drop non-numeric metrics thru Telegraf
Reduce Health exception (when API server response is nil)
Add 'Computer' dimension to all telemetry (internal use)
Support for specifiying HTTP & HTTPS Proxy for outbound/egress (applicable only for non-AKS clusters)
Move to rbac.authorization.k8s.io/v1 for ClusterRole & ClusterRoleBinding
Move to apiextensions.k8s.io/v1 for Health CRD

• Add support for Rate limiting
• Add support for Container Run-time Interface compatible container runtime(s) like CRI-O, ContainerD etc.

Introduce Versioning for monitors

Collection of GPU metrics as InsightsMetrics
Enable config map settings to enable collection of 'Normal' kube events
Fix kubehealth exceptions to handle empty/nil kube api responses
Get resource limits for health and MDM from kubelet instead of kube api
Bug fix for windows node image collection where image name contains multiple slashes
Exclude ARO master node for data collection
Telemetry for kube events flushed
Changes to support msi for mdm if service principal doesnt exist
Changes for AKS telemetry to ping ods endpoint first and then network check
KubeEvents bug fix for KubeEventType