Container host metrics
Host metrics for containerized systems are fully supported.
Container limits
For container host metrics to be accurate, limits need to be set for every container. This means, configuring your container to have a limited number of CPUs and memory allocated to it. Without these limits the container reports the maximum possible value of these metrics, resulting in the host reporting Terabytes of available memory for example. A container without swap configured, or unsupported on the host system, will report a 0
value. For more information on how to limit your container's CPU and memory, please read the documentation on:
Supported metrics
On systems that expose the /sys/fs
virtual file system the following metrics are supported.
Metric | Description |
---|---|
CPU usage | User and system in percentages. Read more about CPU metrics in our academy article. |
Load average | 1 minute load average on the host. |
Memory usage | Available, free and used memory. Also includes swap total and swap used. |
Disk usage | Percentage of every disk used. |
Disk IO | Throughput of data read from and written to every disk. |
Network traffic | Throughput of data received and transmitted through every network interface. |
About CPU metrics
In short:
- AppSignal reports the same metrics as the
docker stats
command, but reported as an average on a minutely basis. - The reported CPU usage can go above 100%. 100% is equal to 1 full virtual CPU core.
- Container limits are recommended to properly balance the available CPU resources between containers.
- When multiple containers on the same host system share CPU resources, even with container limits, containers that use more CPU can negatively impact the reported CPU usage for other containers.
The container CPU metrics work a little differently than normal CPU metrics. They are affected by container limits and how many other containers are active on the system. This is because they share the container system's resources.
Unlike normal host metrics the CPU usage can go above 100%. This is because a container can have more than 1 virtual CPU core available to it. Every full core amounts to 100%. If a container has three cores for example, the maximum is 300%.
Other containers on the system can impact the total CPU time available to the container. Limits will help maintain a better balance of CPU usage between containers so the averages fluctuate less depending on busy neighbors. If all containers share all the host system's CPU resources, the usage on one host can negatively impact the available virtual CPU cores on another.
For example:
We have a Docker system on our host machine that has 3 virtual CPU cores available. We then start containers A, B and C without any specific CPU limits.
- In the scenario of container A maxing out its available CPU cores, and the other containers B and C idling at 0%, it will be reported as 300% CPU usage for container A. Which is the total CPU time available on the host's container system.
- When the three containers are all maxing out their available CPU cores it will be reported as 100% CPU usage for every container, as the resources are shared amongst them evenly.