UPGRADE_DISCOVERY_TIMEOUT is the amount of time given to discover an upgrade is
happening using kapp app-change ls -a cf.
Our SLO is defined as:
95% (X) of GET Requests to a Route succeed in less than 100 (Y) milliseconds.
The SLI used to measure our SLO is request latency.
Request Latency is measured by the following:
Given an app 'A' deployed on the platform with route 'r', the SLI times how long it takes to make an HTTP Get Request to 'r' for 'A' and receive a response.
CF_APP_DOMAIN is the app domain. This is currently used to map new routes to test control plane uptime.
DATA_PLANE_APP_NAME: Name of the app.
X: DATA_PLANE_SLO_PERCENTAGE Y: DATA_PLANE_SLO_MAX_REQUEST_LATENCY r: DATA_PLANE_SLI_APP_ROUTE_URL
Our SLO is defined as:
95% (X) of routes that get mapped become available in less than 15 (Y) seconds.
- Every 5 seconds, map a route
- Sleep for propagation time (Y)
- Send requests to route for 30 seconds (U), record their latency and response code
- If greater than 95% (Z) of those requests had a non-200 response code, or exceeded the latency SLO (W), consider that route to be a failure
- If greater than 95% (X) of routes are failures, fail the test
CONTROL_PLANE_SLO_PERCENTAGE (X): Percentage of routes that we expect to get mapped
and become available in less than number of seconds defined by
CONTROL_PLANE_SLO_MAX_ROUTE_PROPAGATION_TIME. Defaults to 95%.
CONTROL_PLANE_SLO_MAX_ROUTE_PROPAGATION_TIME (Y): Time we wait before seeing if a route is live, defaults to 15 seconds.
CONTROL_PLANE_SLO_DATA_PLANE_AVAILABILITY_PERCENTAGE (Z): Percentage of routes
that succeed in less than the number of seconds defined by
CONTROL_PLANE_SLO_DATA_PLANE_AVAILABILITY_PERCENTAGE. Defaults to 95%.
CONTROL_PLANE_SLO_DATA_PLANE_MAX_REQUEST_LATENCY (W): Max response time from mapped route, defaults to 200ms.
CONTROL_PLANE_SLO_SAMPLE_CAPTURE_TIME (U): How often we send a request to a route. Defaults to 10 seconds.
CONTROL_PLANE_APP_NAME: Name of the app to map routes to. Defaults to
upgrade-control-plane-sli.