As of August 26, 2024, you can no longer create new monitors using legacy runtimes on public or private locations.
On October 22, 2024, we will end of life the containerized private minion (CPM) and the legacy synthetics runtime versions it supports. Please review our recommended migration steps to avoid degradation of your private location monitors.
The CPM can operate in a Docker container system environment or a Kubernetes container orchestration system environment. The CPM will auto-detect its environment to select the appropriate operating mode.
General private minion features
Because the CPM operates as a container instead of a virtual machine, it delivers many features:
Ability to leverage a Docker container as a sandbox environment
Customizable monitor check timeout
Custom provided modules for scripted monitor types
Kubernetes-specific features
Also, the CPM delivers the following features in a Kubernetes environment:
Integrates with the Kubernetes API to delegate runtime lifecycle management to Kubernetes
Does not require privileged access to the Docker socket
Supports hosted and on-premise Kubernetes clusters
Supports various container engines such as Docker and Containerd
Deployable via Helm charts as well as configuration YAMLs
Allows job (ping vs. non-ping checks) based resource allocation for optimum resource management
Observability offered via the New Relic Kubernetes cluster explorer
System requirements and compatibility
To host CPMs, your system must meet the minimum requirements for the chosen system environment.
Advertencia
Do not modify any CPM files. New Relic is not liable for any modifications you
make. For more information, contact your account representative or a New Relic
technical sales rep.
Measured by monitoring write throughput for https://newrelic.com on an AWS EC2 m5.xlarge instance with: AL2, gp2 storage class, 50 GiB root volume, a default Docker CPM install, and 1 monitor at a time set to 1-minute frequency. Efficiency gains are expected with multiple monitors running. These values may be higher or lower than your own depending on many factors.
Baseline for Docker CPM is 0.5 IOPS with no monitor jobs running.
The Docker CPM does not support Docker Engine 26.0 or higher due to breaking changes. Customers seeking Docker 26+ support should migrate to Synthetics Job Manager
Advertencia
The Docker CPM is not designed for use with container orchestrators like AWS ECS, Docker Swarm, Apache Mesos, Azure Container Instances, etc. Running the Docker CPM in a container orchestrator will result in unexpected issues because it is itself a container orchestrator. If you're using container orchestration, see our Kubernetes CPM requirements.
Compatibility for
Requirements
Operating system
Linux kernel: 3.10 or higher macOS: 10.11 or higher
Linux containers, including the containerized private minion, only run on Linux K8s nodes.
Processor
A modern, multi-core CPU
Minion pod
CPU (vCPU/Core): 0.5 up to 0.75 Memory: 800Mi up to 1600Mi
Resources allocated to a Minion pod are user configurable.
Runner pod
CPU (vCPU/Core): 0.5 up to 1 Memory: 1250Mi up to 3000Mi
For a scripted API check, 1250Mi will be requested with a limit of 2500Mi.
For a simple browser or scripted browser check, 2000Mi will be requested with a limit of 3000Mi.
Additional considerations:
Resources allocated to a runner pod are not user configurable.
The maximum limit-request resource ratio for both CPU and memory is 2.
Disk size
Root volume: A minimum of 50 GiB (node + PV) Persistent volume (PV): A minimum of 20 GiB
If a ReadWriteOnce (RWO) PV is provided to the minion, an implicit node affinity will be established to ensure the minion and the runner containers are scheduled on the same node. This is required to allow the minion and the associated runners access to the PV, as an RWO PV can be accessed only by a single node in the cluster.
Measured by monitoring write throughput for https://newrelic.com on an AWS EKS 1.21 cluster backed by one EC2 m5.xlarge node running with: gp2 storage class, 50 GiB root volume, 20 GiB PV/PVC with RWO access mode, a default Kubernetes CPM install via Helm, and 1 monitor at a time set to 1-minute frequency. Efficiency gains are expected with multiple monitors running. These values may be higher or lower than your own depending on many factors.
Baseline for Kubernetes CPM is 3.0 IOPS with no monitor jobs running.
Disk file system
NFSv4.1 or higher (if using NFS)
Kubernetes version
We recommend that your Kubernetes cluster supports Kubernetes v1.15.
Advertencia
For Kubernetes v1.21 or newer, use minion release v3.0.61 or newer.
After August 2022, we're discontinuing support for several capabilities, including our original private minion Helm chart URL. For more details, including how you can easily prepare for this transition, see our Support Forum post.
To view versions, dependencies, default values for how many runner pods start with each minion, the Persistent volume access mode, and more, please see Show help and examples below.
Private location key
Before launching CPMs, you must have a private location key. Your CPM uses the key to authenticate against New Relic and run monitors associated with that private location.
In the Private locations index, locate the private location you want your CPM to be assigned to.
Note the key associated with the private location with the key
icon.
Sandboxing and Docker dependencies
Sandboxing and Docker dependencies are applicable to the CPM in a Docker container system environment.
The CPM runs in Docker and is able to leverage Docker as a sandboxing technology. This ensures complete isolation of the monitor execution, which improves security, reliability, and repeatability. Every time a scripted or browser monitor is executed, the CPM creates a brand new Docker container to run it in called a runner.
The minion container needs to be configured to communicate with the Docker engine in order to spawn additional runner containers. Each spawned container is then dedicated to run a check associated with the synthetic monitor running on the private location the minion container is associated with.
There are two crucial dependencies at launch. To enable sandboxing, ensure that:
Your writable and executable directory is mounted at /tmp. The writable directory can be any directory you want the CPM to write into, but New Relic recommends the system's own /tmp to make things easy.
Your writable Docker UNIX socket is mounted at /var/run/docker.sock or DOCKER_HOSTenvironment variable. For more information, see Docker's Daemon socket option.
Advertencia
Core count on the host determines how many runner containers the CPM can run concurrently on the host. Since memory requirements are scaled to the expected count of runner containers, we recommend not running multiple CPMs on the same host to avoid resource contention.
Both installing and updating the CPM use the same command to pull the latest Docker image from the Quay.io repository where the CPM Docker image is hosted. Go to quay.io/repository/newrelic/synthetics-minion for a list of all the releases.
Unless hosting the images in a local image repository, connections to either quay.io or docker.io will need to be allowed through your firewall in order for Docker to pull the synthetics-minion and synthetics-minion-runner images. The "runner" image is pulled automatically on startup of the synthetics-minion container. See Docker environment configuration and Kubernetes environment configuration for details on how to set a local repository and the runner registry endpoint.
Start the CPM
To start the CPM, follow the applicable Docker or Kubernetes instructions.
Run the appropriate script for your system. Tailor the common defaults for /tmp and /var/run/docker.sock in the following examples to match your system.
When a message similar to Synthetics Minion is ready and servicing location YOUR_PRIVATE_LOCATION_LABEL appears, your CPM is up and ready to run monitors assigned to that location.
EOL NOTICE
After August 2022, we're discontinuing support for several capabilities, including our original private minion Helm chart URL. For more details, including how you can easily prepare for this transition, see our Support Forum post.
Once the status attribute of each pod is shown as running, your CPM is up and ready to run monitors assigned to your private location.
Stop or delete the CPM
On a Docker container system environment, use the Docker stop procedure to stop the CPM from running. On a Kubernetes container orchestration system environment, use the Kubernetes delete procedure to stop the CPM from running.
For a CPM in the Kubernetes container orchestration system environment, the following Helm show commands can be used to view the chart.yaml and the values.yaml, respectively:
bash
$
helm show chart YOUR_REPO_NAME/synthetics-minion
bash
$
helm show values YOUR_REPO_NAME/synthetics-minion
Show license information
To show the licensing information for the open source software that we use in the CPM, run the LICENSE command.
Run this command to view license information for CPM versions 2.2.27 or higher:
bash
$
docker run quay.io/newrelic/synthetics-minion:latest LICENSE
Some of our open-source software is listed under multiple software licenses, and in that case we have listed the license we've chosen to use. Our license information is also available in the our licenses documentation.
Configure CPM
You can configure the containerized private minion with custom node modules, preserve data between launches, use environment variables, and more. For more information, see CPM configuration.
Networks
For both Docker and Kubernetes, the CPM and its runner containers will inherit network settings from the host. For an example of this on a Docker container system environment, see the Docker site.
A new bridge network is created for each runner container. This means networking command options like --network and --dns passed to the CPM container at launch (such as through Docker run commands on a Docker container system environment) are not inherited or used by the runner containers.
When these networks are created, they pull from the default IP address pool configured for daemon. For an example of this on a Docker container system environment, see the Docker site.
Typically, the runner network is removed after the check is completed. However, if a CPM exits while a check is still running, or exits in another unexpected circumstance, these networks may get orphaned. This can potentially use up IP address space that is available to the Docker daemon.
If this happens, you may see INTERNAL ENGINE ERROR code: 31 entries in your CPM logging when trying to create a new runner container. To clean these up in Docker container system environments only, run docker network prune.
Security, sandboxing, and running as non-root
By default, the software running inside a CPM is executed with root user privileges. This is suitable for most scenarios, as the execution is sandboxed.
In a Docker container system environment: To change the default AppArmor profile used by containers that CPM spawns to run monitors, see the environment variableMINION_RUNNER_APPARMOR (CPM version 3.0.3 or higher) or MINION_DOCKER_RUNNER_APPARMOR (CPM version up to v3.0.2).
To run the CPM as a non-root user, additional steps are required:
Verify that my_user has read/write permissions to all the directories and volumes passed to CPM. To set these permission, use the chmod command.
Get the uid of my_user for use in the run command: id -u my_user.
Once these conditions are met, use the option "-u <uid>:<gid>" when launching CPM:
bash
$
docker run ... -u1002...
OR
bash
$
docker run ... -u1002-eDOCKER_HOST=http://localhost:2375 ...
Docker image repository
A single CPM Docker image serves both the Docker container system environment and Kubernetes container orchestration system environment. The Docker image is hosted on quay.io. To make sure your Docker image is up-to-date, see the quay.io newrelic/synthetics-minion repository.
Additional considerations for CPM connection
Connection
Description
CPMs without Internet access
A CPM can operate without access to the internet, but with some exceptions. The public internet health check can be disabled using the environment variables named MINION_NETWORK_HEALTHCHECK_DISABLED for a Docker container system environment or synthetics.minionNetworkHealthCheckDisabled for a Kubernetes container orchestration system environment. The CPM needs to be able to contact the "synthetics-horde.nr-data.net" domain. This is necessary for it to report data to New Relic and to receive monitors to execute. Ask your network administration if this is a problem and how to set up exceptions.
Communicate with synthetics via a proxy
To set up communication with New Relic by proxy, use the environment variables named MINION_API_PROXY*.
Arguments passed at launch
This applies to a Docker container environment only. Arguments passed to the CPM container at launch do not get passed on to the containers spawned by the CPM. Docker has no concept of "inheritance" or a "hierarchy" of containers, and we don't copy the configuration that is passed from CPM to the monitor-running containers. The only shared configuration between them is the one set at the Docker daemon level.