Check which service signals a deploy compared

Open the deployment from the environment’s Deploys list and inspect Signals. Its service mapping can explain why a configured metric is absent from this deployment.

Read the values and check task coverage#

Signals lists the value of each signal the deploy compares in the hour before the deploy started and at each check, and marks a regressed value in red. A deploy starts when its rollout begins, which can be well before it lands while the rollout moves traffic: when its GitHub deployment is created, or when its job starts after an approval or wait timer. For an AWS run, or a wait with no later status, that hour ends no more than an hour before landing. When Proofline holds no start time for the deploy, that hour ends at landing instead. Point to a value under Hour before or Day before to see the minutes it covers. Once the 1-day check has read, Day before shows the value it compared with. Without a signal reading, the page leaves this section out.

A signal that sums or counts, such as requests or 5xx responses, shows each value per second, such as 0.0267/s, because the windows differ in length. The checks compare these rates, not each window's total, so a reading can regress while its window's total is smaller than the hour before's. A mean, minimum or maximum, such as latency, shows the value in the unit the signal measures, such as 212 ms, and the checks compare it as read. A signal's Minimum change is in the same unit as its values.

A deploy compares the signals of the services it changes. When the repository's Deployment tasks map the deploy's GitHub deployment task to Cloud Run services, the deploy compares each Cloud Monitoring signal whose service_name resource label names one of those services where the map places it: in the signal's project and, when the signal's location resource label names a location, in that location. A signal with no location label matches the service wherever the map places it in that project. The deploy also compares each signal that names no service. It leaves out a Cloud Monitoring signal of another service, or of a service of the same name in another project or location, and a CloudWatch signal with an ECS ServiceName dimension, and Signals does not list them. A left-out signal does not make a check incomplete. When they map the task to ECS services, the deploy's checks read the running images of those services only, and the deploy compares each CloudWatch signal whose ServiceName dimension names one of those services where the map places it: in the signal's account and region and, when the signal's ClusterName dimension names a cluster, in that cluster. A signal with no ClusterName dimension matches the service wherever the map places it in that account and region. The deploy also compares each signal that names no service. It leaves out a CloudWatch signal of another service, or of a service of the same name in another account, region or cluster, and a Cloud Monitoring signal whose service_name resource label names a service.

A deploy of a GitHub deployment task compares no signal when Proofline cannot tell which services it changed. That happens when the repository's Deployment tasks leave out the deploy's task while they map another task or the repository has deployed more than one task, and when they map the task only to services the environment does not deliver to. Its checks read no signal and record a gap, so the page leaves Signals out and the deploy counts as Incomplete. For a task the map leaves out, an administrator sees Deployment task not in the task map with Map deployment tasks. Every other deploy, including each deploy of a repository that has deployed one task and maps none, compares every signal the environment watches.

Search help

Tab to a result. Enter to open. Escape to close.