How to run the VM Attribution mock

The VM Attribution mock

The VM Attribution mock simulates a hypervisor Prometheus exporter that reports which which fraction of a baremetal host’s usage is attributed to each of the virtual machines running on it. It serves a single metric on the /metrics endpoint:

vm_attribution_coefficient{name="prometheus-mock-bm-1",resource_consumer_id="/qemu.slice/120.scope"} 0.1

The coefficients always sum to less than one, leaving part of the host usage unattributed to any VM (ie. system overhead).

Simulated values

Like the SNMP mock, the values are not random. The attribution coefficient drifts by at most a fixed variance per scrape and stays between a lower and an upper bound. The reported coefficient is that weight relative to the total.

The mock is configured with environment variables on its container (see dev/k8s/vm-attribution-mock-deployment.yaml):

Variable	Default	Description
`MOCK_PROMETHEUS_HOST_NAME`	`prometheus-mock-bm-1`	Host reported in the `name` label.
`MOCK_PROMETHEUS_VM_IDS`	`120,121`	Comma separated VM IDs to report.
`MOCK_PROMETHEUS_PORT`	`9090`	Port the HTTP server listens on.
`MOCK_PROMETHEUS_WEIGHT_MIN`	`0.2`	Lower bound of a VM weight.
`MOCK_PROMETHEUS_WEIGHT_MAX`	`1.0`	Upper bound of a VM weight.
`MOCK_PROMETHEUS_WEIGHT_START`	`0.6`	Weight used on the first scrape.
`MOCK_PROMETHEUS_WEIGHT_VARIANCE`	`0.02`	Maximum change per scrape, in either direction.

Manual installation

The mock is part of the local development environment. After deploying Chantico (which is deployed by make cluster-configure), you can then deploy the mocks with:

make docker-pull-mocks
make cluster-mocks

This deploys both the mock deployment and the service for both the VM Attribution mock and the SNMP mock (see how to run the SNMP mock).

Querying the mock

The service is exposed on NodePort 30909, which the development kind cluster maps to host port 19091:

curl http://localhost:19091/metrics

Repeated scrapes should show the coefficients changing only slightly.