--- title: Windows Containers Issues description: Datadog, the leading service for cloud-scale monitoring. breadcrumbs: Docs > Agent > Agent Troubleshooting > Windows Containers Issues --- # Windows Containers Issues This page describes known and open issues for Containerized Windows Applications Monitoring. ## Common issues{% #common-issues %} Containerized Windows Applications Monitoring requires Datadog Agent 7.19+. The supported OS versions are: - Windows Server 2019 (LTSC / 1809) - Windows Server 2019 1909 (until Agent 7.39, no longer supported by Microsoft) - Windows Server 2019 2004 or 20H1 (until Agent 7.39, no longer supported by Microsoft) - Windows Server 2019 20H2 (Agent 7.33 to 7.39, no longer supported by Microsoft) - Windows Server 2022 LTSC (Agent >=7.34) Hyper-V isolation mode is not supported. Host metrics for disk, IO, and network are disabled. They are not supported by Windows Server, hence the Agent Checks are disabled by default. ## Docker issues{% #docker-issues %} Live processes do not appear in containers (except for the Datadog Agent). ## Kubernetes issues{% #kubernetes-issues %} Live processes do not appear in containers (except for the Datadog Agent). ### Mixed clusters (Linux + Windows){% #mixed-clusters-linux--windows %} The recommended way of deploying the Datadog Agent on a mixed cluster is to perform two installations of the Helm chart with different `targetSystem`s. The Datadog Agent uses a `nodeSelector` to automatically select Linux or Windows nodes based on `targetSystem`. However, this is not the case for Kube State Metrics (which is installed by default), leading to situations where Kube State Metrics cannot be scheduled on Windows nodes. Three options are available to avoid this issue: - Taint your Windows nodes. On Windows, the Agent always allows the `node.kubernetes.io/os=windows:NoSchedule` taint. - Set Kube State Metrics node selector through Datatog Helm chart `values.yaml`: ``` kube-state-metrics: nodeSelector: beta.kubernetes.io/os: linux // Kubernetes < 1.14 kubernetes.io/os: linux // Kubernetes >= 1.14 ``` - Deploy Kube State Metrics yourself separately by setting `datadog.kubeStateMetricsEnabled` to `false`. **Note**: When using two Datadog installations (one with `targetSystem: linux`, one with `targetSystem: windows`), make sure the second one has `datadog.kubeStateMetricsEnabled` set to `false` to avoid deploying two instances of Kube State Metrics. Some metrics are not available for Windows deployments. See available metrics. #### Mixed clusters with the Datadog Cluster Agent{% #mixed-clusters-with-the-datadog-cluster-agent %} With Cluster Agent v1.18+, a configuration with mixed clusters is supported by the Datadog Cluster Agent. Use the following `values.yaml` file to configure communication between Agents deployed on Windows nodes and the Cluster Agent. ```yaml targetSystem: windows existingClusterAgent: join: true serviceName: "" # from the first Datadog Helm chart tokenSecretName: "" # from the first Datadog Helm chart # Disable datadogMetrics deployment since it should have been already deployed with the first chart. datadog-crds: crds: datadogMetrics: false # Disable kube-state-metrics deployment datadog: kubeStateMetricsEnabled: false ``` #### Limited configuration options for Windows deployments{% #limited-configuration-options-for-windows-deployments %} Some configuration options are not available on Windows. The following is a list of **unsupported** options: | Parameter | Reason | | ---------------------------------------- | ------------------------------------------------ | | `datadog.dogstatsd.useHostPID` | Host PID not supported on Windows containers | | `datadog.dogstatsd.useSocketVolume` | Unix sockets not supported on Windows | | `datadog.dogstatsd.socketPath` | Unix sockets not supported on Windows | | `datadog.processAgent.processCollection` | Unable to access host/other containers processes | | `datadog.systemProbe.seccomp` | System probe is not available for Windows | | `datadog.systemProbe.seccompRoot` | System probe is not available for Windows | | `datadog.systemProbe.debugPort` | System probe is not available for Windows | | `datadog.systemProbe.enableConntrack` | System probe is not available for Windows | | `datadog.systemProbe.bpfDebug` | System probe is not available for Windows | | `datadog.systemProbe.apparmor` | System probe is not available for Windows | | `agents.useHostNetwork` | Host network not supported by Windows Containers | ### HostPort for APM or DogStatsD{% #hostport-for-apm-or-dogstatsd %} `HostPort` is partially supported on Kubernetes, depending on the underlying OS version and CNI plugin. Requirements to have `HostPort` working are the following: - Windows Server version must be >= 1909 - CNI plugin must support `portMappings` capability Currently, at least two CNI plugins support this capability: - Official `win-bridge` plugin (version >= 0.8.6) - used by GKE - Azure CNI Plugin - used by AKS If your setup does not meet these requirements, APM and DogStatsD will only work when pod-to-pod networking is configured between the Tracer and the Agent. ### Kubelet check{% #kubelet-check %} Depending on your Kubernetes version, some Kubelet metrics might not be available (or Kubelet check might timeout). For optimal experience, please use any of the following with Datadog Agent v7.19.2+: - Kubelet v1.16.13+ (v1.16.11+ on GKE) - Kubelet v1.17.9+ (v1.17.6+ on GKE) - Kubelet v1.18.6+ - Kubelet v1.19+ ### Limited metrics for Windows deployments{% #limited-metrics-for-windows-deployments %} The following `kubernetes.*` metrics are available for Windows containers: - `kubernetes.cpu.usage.total` - `kubernetes.containers.restarts` - `kubernetes.containers.running` - `kubernetes.cpu.capacity` - `kubernetes.ephemeral_storage.usage` - `kubernetes.kubelet.container.log_filesystem.used_bytes` - `kubernetes.kubelet.network_plugin.latency.count` - `kubernetes.kubelet.network_plugin.latency.quantile` - `kubernetes.kubelet.network_plugin.latency.sum` - `kubernetes.kubelet.runtime.errors` - `kubernetes.kubelet.runtime.operations` - `kubernetes.memory.capacity` - `kubernetes.pods.running` - `kubernetes.rest.client.latency.count` - `kubernetes.rest.client.latency.sum` - `kubernetes.rest.client.requests` - `kubernetes.network.tx_bytes` - `kubernetes.network.rx_bytes` - `kubernetes.cpu.usage.total` - `kubernetes.memory.working_set` - `kubernetes.filesystem.usage` - `kubernetes.filesystem.usage_pct`