Find changes to a resource an alert identified
In an alert investigation, open Summary and inspect the reasons shown under Where to look. The alert can point to particular services, resources and revisions among the candidates.
Match the alert-specific reasons#
Five reasons appear only in some alert investigations:
- service the alert counted: the lead changed the configuration of a Cloud Run service that the Cloud Monitoring alert's metric counted, or of one of that service's revisions. It ranks above the other leads with the same timing. In Slack and pull request replies, such a lead reads "Configuration change to a service the alert counted".
- resource the alert named: the lead changed the configuration of the resource that the alert itself names. For a Cloud Monitoring alert, that is the Cloud Run service or Cloud SQL instance the alert names, in the alert's project: the service's settings or one of its revisions, or the instance's settings, its database flags, its authorized networks or its server CA certificate. A change to a service the alert also counted reads service the alert counted instead. For a CloudWatch alarm, it is the one resource that all of the alarm's metrics name, in the alarm's account and region: an ECS service or cluster, an RDS instance or cluster, a Lambda function, an SQS queue, an SNS topic, a load balancer or an ACM certificate. A composite alarm, and a metric math alarm whose metrics name more than one resource, name none. For an Azure Monitor alert, it is the resource the alert targets, or the first one when it targets several; a subscription or resource group is not one. It ranks above the other leads with the same timing. In Slack and pull request replies, such a lead reads "Configuration change to the resource the alert named".
- resource an alert label named: a value of one of the Cloud Monitoring alert's metric labels names an asset in the alert's project, such as the instance template that a log-based metric counted, and the lead is that asset being added or removed. Only the asset's additions and removals get this reason. A label value that matches no asset, or more than one, names nothing. The lead ranks above the other leads with the same timing. In Slack and pull request replies, such a lead reads "Configuration change to a resource the alert's labels named". A label that names the asset does not establish that adding or removing it caused the alert.
- landed after an earlier incident of this alert: the commit landed after an earlier incident of the same alert opened, so the alert had fired before the change landed. The reason cites that incident's timeline entry. It does not move the lead.
- not serving when read: the lead is the last deploy before the alert or the symptom, but when Proofline read the timeline, Kian observed every serving Cloud Run revision of its service running other commits that were deployed before it. The reason cites those deploys' timeline entries. It records what Kian observed at that read, not that the release was rolled back, and it does not move the lead. When Proofline could not tell what was serving, no lead gets this reason. A workload with a running ECS image never gets it either: the image's commit is the one that built it, and a deploy that built no new image runs one that an earlier commit built. On ECS, a lead without this reason says nothing about whether its deploy was serving. The same words end the lead's line in Slack and pull request replies.