Troubleshooting
Read the health state and fix the common AWS/Azure connection and collection issues.
troubleshootinghealthauth failedthrottledno metrics
Reading the health state
Each integration shows a health badge. Use it to jump straight to the cause.
| Health | Meaning | What to do |
|---|---|---|
| OK | Last poll succeeded. | Nothing — metrics are flowing. |
| Pending | Not polled yet. | Wait one interval, or Test connection. |
| Auth failed | Credentials/permissions rejected. | Check the role trust + ExternalId (AWS) or consent + Monitoring Reader (Azure). |
| Throttled | The provider is rate-limiting us. | Increase the poll interval or narrow the selection; it recovers automatically. |
| Degraded | Transient errors / partial data. | Check the last error on the row; often resolves on the next poll. |
Common symptoms
“Test connection” fails on AWS
- The role's trust policy is missing VerOps's platform account as principal, or the ExternalId doesn't match.
- The permissions policy is missing
cloudwatch:GetMetricData/ListMetrics. - The role ARN was mistyped.
“Test connection” fails on Azure
- Admin consent wasn't granted, or Monitoring Reader isn't assigned on the scope.
- The Directory (tenant) ID or Subscription ID is wrong.
Connected, but no metrics appear
- AWS: the selected namespaces have no active resources, or the region is wrong.
- Azure: no resource IDs were provided (Phase 1 collects by resource id), or the resource emits no metrics in the window.
- Give it one poll interval — the first datapoints arrive after the first successful poll.
Too many series / higher cost than expected
- Tighten the namespace / resource / metric allow-list.
- Increase the poll interval (default 300s).
Note: The platform-side credentials (VerOps's AWS base identity and Azure app) are managed by your VerOps operator. If the recommended flows show “not configured”, ask your operator to finish platform setup, or use the per-account keys / Service Principal fallback.