Host Maps part 2: quick guide
Click here to read Part 1: Introducing Host Maps.
Navigate to the Host Maps by clicking “Infrastructure” → “Host Maps” in the top nav bar.
Your hosts probably have a lot of tags. Some tags are applied automatically by Datadog integrations, and some tags were probably applied by members of your team. Regardless of how the tags were created, you can use any of them to slice and dice your Host Maps.
If some of your hosts are running on AWS, the following AWS-specific tags are available to you right now:
- and any EC2 tags you might use, such as
“Filter by” limits the Host Maps to a specific subset of your infrastructure. Located in the top-left of Host Maps, the filter input bar lets you filter your map by any of your tags, plus the Datadog-provided attributes below. If your filter input bar is empty, then the map displays all hosts that are reporting metrics to Datadog. If you want to focus your attention on just a subset of your hosts, then add filters. Example: if you tag your hosts by the environment they are in, you can filter by “production” to remove hosts in your staging and other environments from the map. If you want to eliminate all but one host role in production, then add that role to the filter, too, the filters will be ANDed together.
Filterable host attributes (automatically provided):
- up : the host is reporting a heartbeat
- down : the host is not reporting a heartbeat
- muted : Datadog alerts are muted for this host
- agent : the host is running the Datadog Agent
- agent_issue : often indicates an integration problem such failed access to a resource
- upgrade_required : the Datadog Agent requires an upgrade
Group hosts by tags
“Group hosts by tags” spatially arranges your hosts into clusters, or groups. Any host in a group shares the tag or tags you group by. A simple example is grouping your hosts by AWS availability zone. If you add a second grouping tag, such as instance type, then the hosts will be further subdivided into groups, first by availability zone and then by instance type, as seen below.
When you’ve identified a host that you want to investigate, click it for details. You will zoom in and see up to six integrations reporting metrics from that host. (If there are more than six integrations, they will be listed under the “Apps” header in the host’s detail pane, as in the screenshot below.) Click the name of an integration, and you will get a condensed dashboard of metrics for that integration. In the screenshot below, we have clicked “system” to get system metrics such as CPU usage, memory usage, disk latency, etc.
Shapes and colors
By default the color of each host (hexagon) is set to represent the percentage of CPU usage on each host, where the color ranges from green (0% utilized) to red (100% utilized). You can select different metrics from the “Color by” selector. The Host Maps can also communicate an additional, optional metric with the size of the hexagon; use the “Size by” selector. In the screenshot below the size of the hexagons is the 15 minute average load, normalized so that machines’ workloads can be compared even if they have different numbers of cores.
Today the “Color by” selector and “Size by” selector contain only CPU-based metrics: load, idle, wait, etc. We will be adding additional metrics in the very near future.
Note that the “% CPU utilized” metric uses the most reliable and up-to-date measurement of CPU utilization, whether it is being reported by the Datadog agent, or directly by AWS, or vSphere.
Data freshness and meaning
Data in the Host Maps is refreshed about once a minute, unless you are continuously interacting with the map. In that case it will not refresh because it can be disorienting to have colors and shapes spontaneously change while you are still investigating. The bottom right of your screen will tell you when data was last updated.
Beyond this quick introduction, we believe that the best way to learn to use Host Maps is to use them yourself—they are fast and intuitive, and we hope you have a lot of fun as you explore your infrastructure.