관리-도구

편집 파일: azure_monitor_cognitive_services.conf

# you can disable an alarm notification by setting the 'to' line to: silent

# --- Cognitive Services - Availability ---

template: am_cognitive_services_availability
       on: azure_monitor.cognitive_services.availability
    class: Availability
     type: Other
component: Cognitive Services
   lookup: average -5m unaligned of availability
    units: percentage
    every: 1m
     warn: $this < (($status >= $WARNING)  ? (99.9) : (99))
     crit: $this < (($status == $CRITICAL) ? (99)   : (90))
    delay: down 5m multiplier 1.5 max 1h
  summary: Cognitive Services availability on ${label:resource_name}
     info: Success rate of Cognitive Services resource ${label:resource_name} \
           in ${label:resource_group} (${label:region})
       to: sysadmin

# --- Cognitive Services - Errors ---

template: am_cognitive_services_server_errors
       on: azure_monitor.cognitive_services.errors
    class: Errors
     type: Other
component: Cognitive Services
   lookup: average -5m unaligned of server
    units: errors/s
    every: 1m
     warn: $this > (($status >= $WARNING)  ? (1) : (5))
     crit: $this > (($status == $CRITICAL) ? (5) : (20))
    delay: down 5m multiplier 1.5 max 1h
  summary: Cognitive Services server errors on ${label:resource_name}
     info: Rate of server errors (5xx) from Cognitive Services resource ${label:resource_name} \
           in ${label:resource_group} (${label:region})
       to: sysadmin

template: am_cognitive_services_client_errors
       on: azure_monitor.cognitive_services.errors
    class: Errors
     type: Other
component: Cognitive Services
   lookup: average -5m unaligned of client
    units: errors/s
    every: 1m
     warn: $this > (($status >= $WARNING)  ? (10) : (50))
    delay: down 5m multiplier 1.5 max 1h
  summary: Cognitive Services client errors on ${label:resource_name}
     info: Rate of client errors (4xx) from Cognitive Services resource ${label:resource_name} \
           in ${label:resource_group} (${label:region}). \
           High client error rates may indicate misconfigured API calls or invalid requests
       to: sysadmin

# --- Cognitive Services - Latency ---

template: am_cognitive_services_latency
       on: azure_monitor.cognitive_services.latency
    class: Latency
     type: Other
component: Cognitive Services
   lookup: average -5m unaligned of average
    units: milliseconds
    every: 1m
     warn: $this > (($status >= $WARNING)  ? (3000) : (5000))
     crit: $this > (($status == $CRITICAL) ? (5000) : (10000))
    delay: down 5m multiplier 1.5 max 1h
  summary: Cognitive Services latency on ${label:resource_name}
     info: Average API latency of Cognitive Services resource ${label:resource_name} \
           in ${label:resource_group} (${label:region})
       to: sysadmin

# --- Cognitive Services - Rate Limiting ---

template: am_cognitive_services_rate_limit
       on: azure_monitor.cognitive_services.rate_limit
    class: Workload
     type: Other
component: Cognitive Services
   lookup: average -5m unaligned of rate_limit
    units: requests/s
    every: 1m
     warn: $this > (($status >= $WARNING)  ? (1) : (5))
     crit: $this > (($status == $CRITICAL) ? (5) : (20))
    delay: down 5m multiplier 1.5 max 1h
  summary: Cognitive Services rate limiting on ${label:resource_name}
     info: Rate of throttled requests due to rate limiting on Cognitive Services resource ${label:resource_name} \
           in ${label:resource_group} (${label:region}). \
           Indicates the service is hitting its API call quota
       to: sysadmin

# --- Cognitive Services - Blocked Calls ---

template: am_cognitive_services_blocked_calls
       on: azure_monitor.cognitive_services.calls
    class: Errors
     type: Other
component: Cognitive Services
   lookup: average -5m unaligned of blocked
    units: calls/s
    every: 1m
     warn: $this > (($status >= $WARNING)  ? (0) : (5))
     crit: $this > (($status == $CRITICAL) ? (5) : (20))
    delay: down 5m multiplier 1.5 max 1h
  summary: Cognitive Services blocked calls on ${label:resource_name}
     info: Rate of blocked API calls on Cognitive Services resource ${label:resource_name} \
           in ${label:resource_group} (${label:region}). \
           Blocked calls indicate authorization or policy violations
       to: sysadmin

# --- Azure OpenAI - Availability ---

template: am_cognitive_services_openai_availability
       on: azure_monitor.cognitive_services.openai_availability
    class: Availability
     type: Other
component: Cognitive Services
   lookup: average -5m unaligned of availability
    units: percentage
    every: 1m
     warn: $this != nan AND $this < (($status >= $WARNING)  ? (99.9) : (99))
     crit: $this != nan AND $this < (($status == $CRITICAL) ? (99)   : (90))
    delay: down 5m multiplier 1.5 max 1h
  summary: Azure OpenAI availability on ${label:resource_name}
     info: Availability rate of Azure OpenAI service on ${label:resource_name} \
           in ${label:resource_group} (${label:region})
       to: sysadmin

# --- Azure OpenAI - Latency ---

template: am_cognitive_services_openai_time_to_response
       on: azure_monitor.cognitive_services.openai_latency
    class: Latency
     type: Other
component: Cognitive Services
   lookup: average -5m unaligned of time_to_response
    units: milliseconds
    every: 1m
     warn: $this != nan AND $this > (($status >= $WARNING)  ? (5000)  : (10000))
     crit: $this != nan AND $this > (($status == $CRITICAL) ? (10000) : (30000))
    delay: down 5m multiplier 1.5 max 1h
  summary: Azure OpenAI time to response on ${label:resource_name}
     info: Average time to response for Azure OpenAI requests on ${label:resource_name} \
           in ${label:resource_group} (${label:region})
       to: sysadmin

template: am_cognitive_services_openai_time_to_first_token
       on: azure_monitor.cognitive_services.openai_latency
    class: Latency
     type: Other
component: Cognitive Services
   lookup: average -5m unaligned of time_to_first_token
    units: milliseconds
    every: 1m
     warn: $this != nan AND $this > (($status >= $WARNING)  ? (3000) : (5000))
     crit: $this != nan AND $this > (($status == $CRITICAL) ? (5000) : (10000))
    delay: down 5m multiplier 1.5 max 1h
  summary: Azure OpenAI time to first token on ${label:resource_name}
     info: Average normalized time to first token for Azure OpenAI requests on ${label:resource_name} \
           in ${label:resource_group} (${label:region})
       to: sysadmin

# --- Azure OpenAI - Provisioned Utilization ---

template: am_cognitive_services_openai_provisioned_utilization
       on: azure_monitor.cognitive_services.openai_provisioned_utilization
    class: Utilization
     type: Other
component: Cognitive Services
   lookup: average -5m unaligned of utilization
    units: percentage
    every: 1m
     warn: $this != nan AND $this > (($status >= $WARNING)  ? (70) : (80))
     crit: $this != nan AND $this > (($status == $CRITICAL) ? (80) : (90))
    delay: down 5m multiplier 1.5 max 1h
  summary: Azure OpenAI provisioned utilization on ${label:resource_name}
     info: Provisioned-managed utilization of Azure OpenAI deployment on ${label:resource_name} \
           in ${label:resource_group} (${label:region}). \
           High utilization means requests may be throttled or rejected
       to: sysadmin

# --- Models - Availability ---

template: am_cognitive_services_model_availability
       on: azure_monitor.cognitive_services.model_availability
    class: Availability
     type: Other
component: Cognitive Services
   lookup: average -5m unaligned of availability
    units: percentage
    every: 1m
     warn: $this != nan AND $this < (($status >= $WARNING)  ? (99.9) : (99))
     crit: $this != nan AND $this < (($status == $CRITICAL) ? (99)   : (90))
    delay: down 5m multiplier 1.5 max 1h
  summary: Model availability on ${label:resource_name}
     info: Model availability rate on Cognitive Services resource ${label:resource_name} \
           in ${label:resource_group} (${label:region})
       to: sysadmin

# --- Models - Latency ---

template: am_cognitive_services_model_time_to_response
       on: azure_monitor.cognitive_services.model_latency
    class: Latency
     type: Other
component: Cognitive Services
   lookup: average -5m unaligned of time_to_response
    units: milliseconds
    every: 1m
     warn: $this != nan AND $this > (($status >= $WARNING)  ? (5000)  : (10000))
     crit: $this != nan AND $this > (($status == $CRITICAL) ? (10000) : (30000))
    delay: down 5m multiplier 1.5 max 1h
  summary: Model time to response on ${label:resource_name}
     info: Average time to response for model requests on Cognitive Services resource ${label:resource_name} \
           in ${label:resource_group} (${label:region})
       to: sysadmin

template: am_cognitive_services_model_time_to_first_token
       on: azure_monitor.cognitive_services.model_latency
    class: Latency
     type: Other
component: Cognitive Services
   lookup: average -5m unaligned of time_to_first_token
    units: milliseconds
    every: 1m
     warn: $this != nan AND $this > (($status >= $WARNING)  ? (3000) : (5000))
     crit: $this != nan AND $this > (($status == $CRITICAL) ? (5000) : (10000))
    delay: down 5m multiplier 1.5 max 1h
  summary: Model time to first token on ${label:resource_name}
     info: Average normalized time to first token for model requests on Cognitive Services resource ${label:resource_name} \
           in ${label:resource_group} (${label:region})
       to: sysadmin

# --- Models - Provisioned Utilization ---

template: am_cognitive_services_model_provisioned_utilization
       on: azure_monitor.cognitive_services.model_provisioned_utilization
    class: Utilization
     type: Other
component: Cognitive Services
   lookup: average -5m unaligned of utilization
    units: percentage
    every: 1m
     warn: $this != nan AND $this > (($status >= $WARNING)  ? (70) : (80))
     crit: $this != nan AND $this > (($status == $CRITICAL) ? (80) : (90))
    delay: down 5m multiplier 1.5 max 1h
  summary: Model provisioned utilization on ${label:resource_name}
     info: Provisioned utilization of model deployment on Cognitive Services resource ${label:resource_name} \
           in ${label:resource_group} (${label:region}). \
           High utilization means requests may be throttled or rejected
       to: sysadmin

# --- Content Safety - Harmful Requests ---

template: am_cognitive_services_harmful_requests
       on: azure_monitor.cognitive_services.content_safety_requests
    class: Errors
     type: Other
component: Cognitive Services
   lookup: average -5m unaligned of harmful
    units: requests/s
    every: 1m
     warn: $this != nan AND $this > (($status >= $WARNING)  ? (1) : (5))
     crit: $this != nan AND $this > (($status == $CRITICAL) ? (5) : (20))
    delay: down 5m multiplier 1.5 max 1h
  summary: Harmful content requests on ${label:resource_name}
     info: Rate of requests flagged as harmful by content safety on Cognitive Services resource ${label:resource_name} \
           in ${label:resource_group} (${label:region}). \
           High rates may indicate abuse or prompt injection attempts
       to: sysadmin

template: am_cognitive_services_blocked_content_requests
       on: azure_monitor.cognitive_services.content_safety_requests
    class: Errors
     type: Other
component: Cognitive Services
   lookup: average -5m unaligned of blocked
    units: requests/s
    every: 1m
     warn: $this != nan AND $this > (($status >= $WARNING)  ? (1) : (10))
    delay: down 5m multiplier 1.5 max 1h
  summary: Blocked content requests on ${label:resource_name}
     info: Rate of requests rejected by content safety on Cognitive Services resource ${label:resource_name} \
           in ${label:resource_group} (${label:region})
       to: sysadmin

# --- Content Safety - Abusive Users ---

template: am_cognitive_services_abusive_users
       on: azure_monitor.cognitive_services.content_safety_abusive_users
    class: Errors
     type: Other
component: Cognitive Services
   lookup: average -5m unaligned of abusive_users
    units: users/s
    every: 1m
     warn: $this != nan AND $this > (($status >= $WARNING) ? (0) : (1))
    delay: down 5m multiplier 1.5 max 1h
  summary: Abusive users detected on ${label:resource_name}
     info: Rate of potentially abusive users detected on Cognitive Services resource ${label:resource_name} \
           in ${label:resource_group} (${label:region})
       to: sysadmin