Docker-Provider

This is a release for Azure AKS monitoring. This is not intended to be used for oms-solution.

• Container logs enrichment optimization
• Get container meta data only for containers in current node (vs cluster before)
• Update fluent bit 0.13.7 => 0.14.4
• This fixes the escaping issue in the container logs
• Mooncake cloud support for agent (AKS only)
• Ability to disable agent telemetry
• Ability to onboard and ingest to mooncake cloud
• Add & populate 'ContainerStatusReason' column to KubePodInventory
• Alertable (custom) metrics (to AzureMonitor - only for AKS clusters)
• Cpuusagenanocores & % metric
• MemoryWorkingsetBytes & % metric
• MemoryRssBytes & % metric
• Podcount by node, phase & namespace metric
• Nodecount metric
• ContainerNodeInventory_CL to fixed type

1. Fix JSON Log Parsing issues in fluent-bit
2. get pods for current node only when calling the kube-api

Updating omsagent in master branch to version 1.8.1-256

This is a release for Azure AKS monitoring. This is not intended to be used for oms-solution.

Fixes:

• Disable Container Image inventory workflow
• Kube_Events memory leak fix for replica-set
• Timeout (30 secs) for outOMS
• Reduce critical lock duration for quicker log processing (for log enrichment)
• Disable OMI based Container Inventory workflow to fluentD based Container Inventory
• Moby support for the new Container Inventory workflow
• Ability to disable environment variables collection by individual container
• Bugfix - No inventory data due to container status(es) not available
• Agent telemetry cpu usage & memory usage (for DaemonSet and ReplicaSet)
• Agent telemetry - log generation rate
• Agent telemetry - container count per node
• Agent telemetry - collect container logs from agent (DaemonSet and ReplicaSet) as AI trace
• Agent telemetry - errors/exceptions for Container Inventory workflow
• Agent telemetry - Container Inventory Heartbeat

This is a release for Azure AKS monitoring. This is not intended to be used for oms-solution.

• Omsagent - 1.8.1.256 (nov 2018 release)
• Persist fluentbit state between container restarts
• Populate 'TimeOfCommand' for agent ingest time for container logs
• Get node cpu usage from cpuusagenanoseconds (and convert to cpuusgaenanocores)
• Container Node Inventory - move to fluentD from OMI
• Mount docker.sock (Daemon set) as /var/run/host
• Liveness probe (Daemon set) - check for omsagent user permissions in docker.sock and update as necessary (required when docker daemon gets restarted)
• Move to fixed type for kubeevents & kubeservices
• Disable collecting ENV for our oms agent container (daemonset & replicaset)
• Disable container inventory collection for 'sandbox' containers & non kubernetes managed containers
• Agent telemetry - ContainerLogsAgentSideLatencyMs
• Agent telemetry - PodCount
• Agent telemetry - ControllerCount
• Agent telemetry - K8S Version
• Agent telemetry - NodeCoreCapacity
• Agent telemetry - NodeMemoryCapacity
• Agent telemetry - KubeEvents (exceptions)
• Agent telemetry - Kubenodes (exceptions)
• Agent telemetry - kubepods (exceptions)
• Agent telemetry - kubeservices (exceptions)
• Agent telemetry - Daemonset , Replicaset as dimensions (bug fix)

Please do not use this private fix (for general users) as this fix ix targeted for a specific customer scenario

• This is preview-release for Azure AKS monitoring. This is not intended to be used for oms-solution.

Fixes:

• Fix for containerID being 00000-00000-00000
• Move from fluentD to fluentbit for container log collection
• Seg fault fixes in json parsing for container inventory & container image inventory
• Telemetry enablement
• Remove ContainerPerf, ContainerServiceLog, ContainerProcess fluentd-->OMI workflows
• Update log level for all fluentD based workflows

• More fixes for CJSON parsing errors

• This is preview-release for Azure AKS monitoring. This is not intended to be used for oms-solution.

Fixes :

• Changes for node lost scenario (roll-up pod & container statuses as Unknown)
• Discover unscheduled pods
• KubeNodeInventory - delimit multiple true node conditions for node status
• UTF Encoding support for container logs
• Container environment variable truncated to 200K
• Handle json parsing errors for OMI provider for docker
• Test mode enablement for ACS-engine testing

This release is for OMS container solution monitoring

Fixes :

• UTF encoding support for Container logs
• Truncate ENV variables size to 200k
• Handlers for json parsing for OMI Provider

*This release is for AKS.

This release is only for AKS monitoring in azure.

Picking up container log from /var/lib/docker/containers/

This release is primarily for kubernetes monitoring in azure.

v1 release for Kubernetes monitoring

Bug fixes:

1. Instead of assuming that container log path is /var/lib/docker/containers/{ContainerID}/{ContainerID}-json.log we do a docker inspect on the container to get the log path
   Refer [PR] (https://github.com/Microsoft/Docker-Provider/pull/80) for further details
2. Decide orchestrator type = Kubernetes based on the existence of environment variable KUBERNETES_SERVICE_HOST instead of the format of the hostname
   Refer PR for further details

Fixes to the installer to address the following things

• Use var/opt/microsoft/docker-cimprov/log/kubernetes_client_log.txt instead of /var/opt/microsoft/omsagent/log/kubernetes_client_log.txt. This accounts for a race condition where the omsagent is installed but /var/opt/microsoft/log directory isnt created yet because of which sometimes there are errors while loading KubernetesApiClient.rb . It doesnt have any impact towards functionality , however it shows up as an error in the omsagent log.
• Clean up properly when run with --remove or --purge
• Set log levels to debug for KubeLogs to decrease log verbosity
• Setup sudo permissions for containerlogtailfilereader which is used for tailing docker container log files

• Fixed exception handling in kube api
• Dedup kube events based on event id and count instead of just event id
• Change in container log collection mechanism. Pick it up from the /var/lib/docker/containers//*.log instead of the docker rest api
• Separate buffers for each data type to improve performance and avoid deadlocks in cases where the collection of one datatype is somehow stuck. This ensures that datatype upload is independent of one another.

Fix:
-Use the agent node to talk to kubernetes api instead of master.

-Kubernetes cluster awareness with at-a-glance hierarchy inventory from Kubernetes cluster to pods
-New Kubernetes events
-Capture custom pod labels and provides custom complex search filters
-Provides container process information
-Container Node Inventory including storage, network, orchestration type, and Docker version