You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
[Azure OpenAI](https://learn.microsoft.com/en-us/azure/ai-foundry/openai/overview) is a fully managed platform that provides access to cutting-edge generative AI models developed by OpenAI, such as GPT, Codex, and Embeddings, through Azure’s secure, enterprise-grade environment. It integrates seamlessly with Azure services like Cognitive Search, Machine Learning, and Logic Apps, as well as external applications and data sources, enabling powerful natural language, code generation, and reasoning capabilities. This integration allows monitoring of key operational and performance metrics such as request volume, token usage, response latency, and error rates, ensuring efficient model utilization and reliable AI-driven application performance.
11
+
[Azure OpenAI](https://learn.microsoft.com/en-us/azure/ai-foundry/openai/overview) is a fully managed platform that provides access to cutting-edge generative AI models developed by OpenAI, such as GPT, Codex, and Embeddings, through Azure’s secure, enterprise-grade environment. It integrates seamlessly with Azure services like Cognitive Search, Machine Learning, and Logic Apps, as well as external applications and data sources, enabling powerful natural language, code generation, and reasoning capabilities. This integration enables the monitoring of key operational and performance metrics, including request volume, token usage, response latency, and error rates, ensuring efficient model utilization and reliable AI-driven application performance.
12
12
13
13
## Log and metric types
14
14
@@ -23,12 +23,12 @@ For more information on supported metrics, refer to [Azure documentation](https:
23
23
24
24
## Setup
25
25
26
-
Azure services sends monitoring data to Azure Monitor, which can then [stream data to Eventhub](https://learn.microsoft.com/en-us/azure/azure-monitor/essentials/stream-monitoring-data-event-hubs). Sumo Logic supports:
26
+
Azure services send monitoring data to Azure Monitor, which can then [stream data to Event Hub](https://learn.microsoft.com/en-us/azure/azure-monitor/essentials/stream-monitoring-data-event-hubs). Sumo Logic supports:
27
27
28
28
* Logs collection from [Azure Monitor](https://docs.microsoft.com/en-us/azure/monitoring-and-diagnostics/monitoring-get-started) using our [Azure Event Hubs source](/docs/send-data/collect-from-other-data-sources/azure-monitoring/ms-azure-event-hubs-source/).
29
29
* Metrics collection using our [Azure Metrics Source](/docs/send-data/hosted-collectors/microsoft-source/azure-metrics-source).
30
30
31
-
You must explicitly enable diagnostic settings for each OpenAI resource you want to monitor. You can forward logs to the same Event hub provided they satisfy the limitations and permissions as described [here](https://learn.microsoft.com/en-us/azure/azure-monitor/essentials/diagnostic-settings?tabs=portal#destination-limitations).
31
+
You must explicitly enable diagnostic settings for each OpenAI resource you want to monitor. You can forward logs to the same Event Hub, provided they satisfy the limitations and permissions as described [here](https://learn.microsoft.com/en-us/azure/azure-monitor/essentials/diagnostic-settings?tabs=portal#destination-limitations).
32
32
33
33
When you configure the event hubs source or HTTP source, plan your source category to ease the querying process. A hierarchical approach allows you to make use of wildcards. For example: `Azure/OpenAI/Logs`, `Azure/OpenAI/Metrics`.
34
34
@@ -49,7 +49,7 @@ In this section, you will configure a pipeline for shipping diagnostic logs from
49
49
#### Diagnostic logs
50
50
51
51
1. To set up the Azure Event Hubs source in Sumo Logic, refer to the [Azure Event Hubs Source for Logs](/docs/send-data/collect-from-other-data-sources/azure-monitoring/ms-azure-event-hubs-source/).
52
-
1. To create the diagnostic settings in Azure portal, refer to the [Azure documentation](https://learn.microsoft.com/en-us/azure/azure-monitor/essentials/diagnostic-settings?tabs=portal#create-diagnostic-settings). Perform the steps below for each Azure Event Hubs namespace that you want to monitor.
52
+
1. To create the diagnostic settings in the Azure portal, refer to the [Azure documentation](https://learn.microsoft.com/en-us/azure/azure-monitor/essentials/diagnostic-settings?tabs=portal#create-diagnostic-settings). Perform the steps below for each Azure Event Hubs namespace that you want to monitor.
53
53
1. Choose `Stream to an event hub` as the destination.
54
54
1. Select `allLogs`.
55
55
1. Use the Event Hub namespace and Event Hub name configured in the previous step in the destination details section. You can use the default policy `RootManageSharedAccessKey` as the policy name.<br/><img src={useBaseUrl('img/send-data/azure-openai-logs.png')} alt="Azure OpenAI logs" style={{border: '1px solid gray'}} width="800" />
@@ -76,8 +76,8 @@ As part of the app installation process, the following fields will be created by
76
76
-`provider_name`. Azure resource provider name (for example, Microsoft.Network).
77
77
-`resource_type`. Azure resource type (for example, storage accounts).
78
78
-`resource_name`. The name of the resource (for example, storage account name).
79
-
-`service_type`. Type of the service that can be accessed with an Azure resource.
80
-
-`service_name`. Services that can be accessed with an Azure resource. (For example, in Azure Container Instances the service is Subscriptions.)
79
+
-`service_type`. The type of service that can be accessed with an Azure resource.
80
+
-`service_name`. Services that can be accessed with an Azure resource. (For example, in Azure Container Instances, the service is Subscriptions.)
81
81
82
82
## Viewing the Azure OpenAI dashboards
83
83
@@ -92,12 +92,12 @@ The **Azure OpenAI - Overview** dashboard provides a high‑level view of the ov
92
92
93
93
### Models
94
94
95
-
The **Azure OpenAI - Models** dashboard enables deep dive into individual model performance, usage, and health. Tracks model availability, request rates, operations, latency, throughput (tokens per second), and usage split by deployment, model name, and resources
95
+
The **Azure OpenAI - Models** dashboard enables a deep dive into individual model performance, usage, and health. Tracks model availability, request rates, operations, latency, throughput (tokens per second), and usage split by deployment, model name, and resources
The **Azure OpenAI - Performance and Latency** dashboard focuses on the responsiveness of Azure OpenAI APIs and models. It tracks time-to-first-byte (TTFB), time-to-response, time-between-tokens for streaming performance, tokens-per-second speed, and time-to-last-byte. Use this dashboard to identify latency bottlenecks across models, deployments, and to compare streaming vs non-streaming performance trends.
100
+
The **Azure OpenAI - Performance and Latency** dashboard focuses on the responsiveness of Azure OpenAI APIs and models. It tracks time-to-first-byte (TTFB), time-to-response, time-between-tokens for streaming performance, tokens-per-second speed, and time-to-last-byte. Use this dashboard to identify latency bottlenecks across models, deployments and to compare streaming vs non-streaming performance trends.
@@ -107,7 +107,7 @@ The **Azure OpenAI - Reliability and Availability** dashboard provides visibilit
107
107
108
108
### Usage and Token Consumption
109
109
110
-
The **Azure OpenAI - Usage and Token Consumption** dashboard provides details on model utilization and token consumption across deployments. The dashboard surfaces prompt tokens (input), generated tokens (output), total tokens processed, and cache match rates. Use this dashboard for cost optimization, and understanding workload trends across different models and regions.
110
+
The **Azure OpenAI - Usage and Token Consumption** dashboard provides details on model utilization and token consumption across deployments. The dashboard surfaces prompt tokens (input), generated tokens (output), total tokens processed, and cache match rates. Use this dashboard for cost optimization and understanding workload trends across different models and regions.
|`Azure OpenAI - Availability`| This alert is triggered when availability of resource drops below 100% . | Count < 100 | Count = 100|
153
-
|`Azure OpenAI - Processed Inference Tokens`| This alert is triggered when inference token consumption crosses the value of 1000000 tokens. | Count > 1000000| Count < = 1000000 |
150
+
| Alert Name | Alert Description and Conditions | Alert Condition | Recover Condition |
151
+
|:--|:--|:--|:--|
152
+
|`Azure OpenAI - Availability`| This alert is triggered when the availability of the resource drops below 100%. | Count < 100 | Count = 100 |
153
+
|`Azure OpenAI - Processed Inference Tokens`| This alert is triggered when inference token consumption crosses the value of 1000000 tokens. | Count > 1000000 | Count < = 1000000 |
154
154
155
155
## Upgrade/Downgrade the Azure OpenAI app (optional)
156
156
@@ -168,4 +168,4 @@ import AppUninstall from '../../reuse/apps/app-uninstall.md';
168
168
169
169
### Metrics collection via Azure Metrics Source
170
170
171
-
To troubleshoot metrics collection via Azure Metrics Source, follow the instructions in [Troubleshooting Azure Metrics Source](/docs/send-data/hosted-collectors/microsoft-source/azure-metrics-source/#troubleshooting).
171
+
To troubleshoot metrics collection via Azure Metrics Source, follow the instructions in [Troubleshooting Azure Metrics Source](/docs/send-data/hosted-collectors/microsoft-source/azure-metrics-source/#troubleshooting).
0 commit comments