Skip to main content
⏳ Estimated read time: 1 min read

Azure OpenAI

StackGen's [Kubernetes agent](../../../../Collecting Telemetry/Install the OpsVerse Agent#VEqXM) can collect metrics from Azure OpenAI.

Add the following YAML snippets to the agent's values.yaml and run the agent based on these [instructions](../../../../Collecting Telemetry/Install the OpsVerse Agent#VEqXM).

info

Replace the placeholders <azure-client-id>, <azure-tenant-id>, <azure-client-secret> in the following YAML with the appropriate values.

azure-metrics-exporter:
enabled: true
env:
DEBUG: "true"
VERBOSE: "true"
AZURE_CLIENT_ID: "<azure-client-id>"
AZURE_TENANT_ID: "<azure-tenant-id>"
AZURE_CLIENT_SECRET: "<azure-client-secret>"

The below job snippet must be added under extraScrapeConfigs of victoria-metrics-agent confing in the agent.

info

Replace the placeholders <azure-subscription-id> in the following YAML with the appropriate values.

victoria-metrics-agent:
extraScrapeConfigs:
- job_name: az-openai-metrics
scrape_interval: 30s
metrics_path: /probe/metrics/list
params:
name:
- az_openai
template:
- '{name}_{metric}'
subscription:
- <azure-subscription-id>
resourceType:
- microsoft.cognitiveservices/accounts
metric:
- RAISystemEvent
- RAITotalRequests
- RAIHarmfulRequests
- RAIRejectedRequests
- SuccessRate
- TotalErrors
- TotalCalls
- SuccessfulCalls
- ServerErrors
- Ratelimit
- DataOut
- DataIn
- ClientErrors
- BlockedCalls
- ProcessedPromptTokens
- TokenTransaction
- GeneratedTokens
- AzureOpenAIRequests
- AzureOpenAIContextTokensCacheMatchRate
- AzureOpenAITimeToResponse
- ActiveTokens
- FineTunedTrainingHours
- Latency
- RAIAbusiveUserCount
- AzureOpenAIProvisionedManagedUtilization
- AzureOpenAIProvisionedManagedUtilizationV2
validateDimensions:
- 'true'
static_configs:
- targets:
- http://devopsnow-agent-azure-metr.devopsnow.svc.cluster.local:8080