Monitoring the health of all the pieces that make up DC/OS is vital to datacenter operators and for troubleshooting hard-to-diagnose bugs. You can monitor the health of your cluster components from the DC/OS UI component health page. The component health page displays information from the system health API, which monitors the core DC/OS components.
DC/OS components are the systemd units that make up the core of DC/OS. These components are monitored by our internal diagnostics utility (
dcos-3dt.service). This utility scans all the DC/OS units, and then exposes an HTTP API on each host. For a complete description of the DC/OS components, see the documentation.
The component health page provides the health status of all DC/OS system components that are running in systemd. You can drill down by health status, host IP address, or specific systemd unit.
Launch the DC/OS UI and navigate to the System -> Components page. You can sort components by health.
You can click on a DC/OS component to view the details, including role, node, and health.
You can debug further by clicking the node to view the component journald (log) output.
Possible health states are unhealthy and healthy. We infer this from codes 0 and 1.
- Healthy All cluster nodes are healthy. The units are loaded and not in the “active” or “inactive” state.
- Unhealthy One or more nodes have issues. The units are not loaded or are in the “active” or “inactive” state.
The system health API has four possible states: 0 - 3, OK; CRITICAL; WARNING; UNKNOWN. Future DC/OS iterations will leverage these codes to give more robust and detailed cluster health state information in the UI.
System health HTTP API endpoint
The system health endpoint is exposed through the DC/OS diagnostics utility on the master nodes:
curl --unix-socket /run/dcos/3dt.sock http://localhost/system/health/v1
Aggregation of the cluster health endpoints is accomplished by the same diagnostics application on the master nodes. You can explore this API further by making a few queries to any master in your cluster:
SSH to your master node:
dcos node ssh --master-proxy --leader
Run this command to open a root session:
sudo su -
Run these commands to get cluster health:
System health by unit:
curl --unix-socket /run/dcos/3dt.sock http://localhost/system/health/v1/units
System health by node:
curl --unix-socket /run/dcos/3dt.sock http://localhost/system/health/v1/nodes
System health report:
curl --unix-socket /run/dcos/3dt.sock http://localhost/system/health/v1/report
The DC/OS user interface uses these aggregation endpoints to generate the data you explore in the system health console.
Misinterpreting System Health by Unit
You can sort system health by systemd unit. However, this search can bring up misleading information as the service itself can be healthy but the node on which it runs is not. This manifests itself as a service showing “healthy” but nodes associated with that service as “unhealthy”. Some people find this behavior confusing.
Missing Cluster Hosts
The system health API relies on Mesos-DNS to know about all the cluster hosts. It finds these hosts by combining a query from
mesos.master A records as well as
leader.mesos:5050/slaves to get the complete list of hosts in the cluster.
This system has a known bug where an agent will not show up in the list returned from
leader.mesos:5050/slaves if the Mesos agent service is not healthy. This means the system health API will not show this host.
If you experience this behavior it’s most likely your Mesos agent service on the missing host is unhealthy.
If you have any problems, you can check if the diagnostics service is running by SSH’ing to the Mesos leading master and checking the systemd status of the diagnostics component (
Here are some recommendations for monitoring a DC/OS cluster. You can use any monitoring tools. The endpoints listed below will help you troubleshoot when issues occur.…Read More
DC/OS cluster nodes generate logs that contain diagnostic and status information for DC/OS core components and DC/OS services.…Read More
DC/OS offers several tools to debug your services when they are stuck in deployment or are not behaving as you expect. This topic discusses how to debug your services using both the DC/OS CLI and the DC/OS web interface.…Read More