-1. Create new policy

-1. Create new policy

-[Environment variables and app settings reference](reference-app-settings.md)

-> * Azure Monitor tools used to used to collect and analyze data

-mms.devlang: na

-## Removing the Defender extension

-This section describes how integration with Microsoft Sentinel can help SOC teams bridge the gap between IT and OT security sectors.

-As a result, CISOs and Security Operations Center (SOC) teams are increasingly responsible for new threats from OT network segments that they traditionally did not manage.

-As a result, SOC solutions and teams are faced with various challenges, including:

-The Defender for IoT workbooks provide guided investigations for OT entities based on open incidents, alert notifications, and activities for OT assets. They also provide a hunting experience across the MITRE ATT&CK® framework for ICS, and are designed to enable analysts, security engineers, and MSSPs to gain situational awareness of OT security posture.

- Wait until the validation state changes to **Pending**. This operation could take a few mins.

-# Configure and access Azure Database for Maria DB audit logs in the Azure CLI

-# Configure and access Azure Database for Maria DB slow query logs by using Azure CLI

-Azure Virtual Network (VNet) is the fundamental building block for your private network in Azure. You can use a VNets to:

-# Connect to and manage Azure databases for MySQL in Azure Purview

-This article outlines how to register an Azure Database for MySQL, and how to authenticate and interact with Azure Database for MySQLs in Azure Purview. For more information about Azure Purview, read the [introductory article](overview.md).

-+ [Image Analysis](cognitive-search-skill-image-analysis.md) that describe images through visual features

-The following screenshot illustrates descriptions and tags for the following embedded images in a PDF. Image analysis detected three images. Other text in the example (including titles, headings, and body text) are text in the source document as well, and therefore were not included in OCR or image processing.

-## Scenario: Embedded images

-When you run OCR over source files that included documents with embedded images, the output is a single string containing all file contents, both text and image-origin text, achieved with this workflow:

-1. The indexer cracks source documents, extracts text and images, and queues each content type for text and image processing.

-1. Images are normalized and passed into enriched documents as a [`"document/normalized_images"`](#get-normalized-images) node.

-1. OCR runs using `"/document/normalized_images"` as input.

-1. [Text Merge](cognitive-search-skill-textmerger.md) runs, combining the text representation of those images with the raw text extracted from the file. Text chunks are consolidated into a single large string.

-1. Indexers reference output field mappings to send skill outputs to fields in a search index.

-The following example skillset creates a "merged_text" field containing the textual content of your document. It also includes the OCRed text from each of the embedded images.

-description: Debug sessions, accessed through the Azure portal, provides an IDE like environment where you can identify and fix errors, validate changes, and push changes to skillsets in the AI enrichment pipeline. Debug sessions is in preview.

-# Debug sessions in Azure Cognitive Search

-Debug sessions is a visual editor that works with an existing skillset in the Azure portal. Within a debug session you can identify and resolve errors, validate changes, and push changes to a production skillset in the AI enrichment pipeline.

-> Debug sessions is a preview feature provided without a service level agreement, and is not recommended for production workloads. For more information, see [Supplemental Terms of Use for Microsoft Azure Previews](https://azure.microsoft.com/support/legal/preview-supplemental-terms/).

->

-## Using debug sessions

-When you start a session, the service creates a copy of the skillset, indexer, and index where a single document is used to test the skillset. Changes made within the session are saved to the session. The changes made within the debug session will not affect the production skillset unless the changes are committed to it. Committing changes will overwrite the production skillset.

-During a debug session you can edit a skillset, inspect, and validate each node in the enrichment tree in the context of a specific document. Once the enrichment pipeline issues are resolved, changes can be saved to the session and committed to the skillset in production.

-## Creating a debug session

-You can create a debug session on the **Overview** page of your search service in the Azure portal. To start a debug session you must have an existing AI enrichment pipeline including; a data source, a skillset, an indexer, and an index. To configure a debug session, you need to name the session, and provide a general-purpose storage account that will be used to cache the skill executions during the indexer run. You will also need to select the indexer that will be running. The indexer has references stored to the data source, skillset, and index. The debug session will default to the first document in the data source or you can specify a document in the data source to step through.

-> :::image type="content" source="media/cognitive-search-debug/debug-session-new.png" alt-text="Creating a debug session":::

-## Debug session features

-The debug session begins by executing the skillset on the selected document. The debug session will record additional metadata associated with each operation in the skillset. The metadata created by the pipeline's skill executions, informs the following set of features that are then used to help identify and fix issues with the skillset.

-## AI Enrichments

-As skills get executed, a tree of enrichments representing the document grows. Using a tree to visualize the outputs of skills or enrichments provides a comprehensive look at all the enrichments performed. You can look across the entire document and inspect each node of the enrichment tree. This perspective makes it easier to shape objects. This format also provides visual cues to the type, path, and contents of each node in the tree.

-## Skill Graph

-The **Skill Graph** view provides a hierarchical, visual representation of the skillset. The graph is a top to bottom representation of the order in which the skills are executed. Skills that are dependent upon the output of other skills will be shown lower in the graph. Skills at the same level in the hierarchy can execute in parallel.

-Selecting a skill in the graph will highlight the skills connected to it, the nodes that create its inputs and the nodes that accept its outputs. Each skill node displays its type, errors or warnings, and execution counts. The **Skill Graph** is where you will select which skill to debug or enhance. When you select a skill, its details will be displayed in the skill details pane to the right of the graph.

-> :::image type="content" source="media/cognitive-search-debug/skills-graph.png" alt-text="Skill Graph":::

-## Skill details

-The skill details pane displays a set of areas for working with a specific skill, when that skill is highlighted in the **Skill Graph**. You can review and edit the details of the skill's settings. The skill's JSON definition is provided. The details of the skill's execution and the errors and warnings are also displayed. The **Skill Settings** tab & **Skill JSON Editor** allow for direct edits to the skill. The [`</>`](#expression-evaluator) opens a window for viewing and editing the expressions of the skills inputs and outputs.

-Nested input controls in the skill settings window can be used to build complex shapes for projections, output field mappings for a complex type field, or an input to a skill. When used with the **Expression evaluator**, nested inputs provide an easy test and validate expression builder.

-## Skill execution history

-A skill can execute multiple times in a skillset for a single document. For example, the OCR skill will execute once for each image extracted from each document. In the skill details pane the **Executions** tab displays the skill's execution history providing a deeper look into each invocation of the skill.

-The execution history enables tracking a specific enrichment back to the skill that generated it. Clicking on a skill input navigates to the skill that generated that input. This allows identification of the root cause of a problem that may manifest in a downstream skill.

-When a potential issue is identified, the execution history displays links to the skills that generated the specific inputs, providing a stack-trace like feature. Clicking on the skill associated with an input, navigates to the skill to fix any bugs or continue to trace the specific issue.

-When building a custom skill or debugging an error with a custom skill, there is the option to generate a request for a skill invocation in the execution history.

-## Enriched Data Structure

-The **Enriched Data Structure** pane shows the document's enrichments through the skillset, detailing the context for each enrichment and the originating skill. The **Expression evaluator** can also be used to view the contents for each enrichment.

-> :::image type="content" source="media/cognitive-search-debug/enriched-data-structure-display.png" alt-text="Enriched Data Structure":::

-## Expression evaluator

-**Expression evaluator** gives a quick peek into the value of any path. It allows for editing the path and testing the results before updating any of the inputs or context for a skill or projection.

-## Errors/Warnings

-This window displays all of the errors and warnings the skillset produces as it is executed against the document in the debug session.

-## Limitations

-Debug sessions work with all generally available data sources and most preview data sources. The MongoDB API (preview) of Cosmos DB are currently not supported.

-Now that you understand the elements of Debug sessions try the tutorial for a hands-on experience.

-> [Explore Debug sessions feature tutorial](./cognitive-search-tutorial-debug-sessions.md)

-| `text` | The text to analyze. |

-In this article, you'll use **Debug sessions** to find and fix 1) a missing input, and 2) output field mappings. The tutorial is all-inclusive. It provides data for you to index (clinical trials data), a Postman collection that creates objects, and instructions for using **Debug sessions** to find and fix problems in the skillset.

-> Debug sessions is a preview feature provided without a service level agreement, and is not recommended for production workloads. For more information, see [Supplemental Terms of Use for Microsoft Azure Previews](https://azure.microsoft.com/support/legal/preview-supplemental-terms/).

-+ An Azure account with an active subscription. [Create an account for free](https://azure.microsoft.com/free/).

-+ An Azure Cognitive Search service. [Create a service](search-create-service-portal.md) or [find an existing service](https://ms.portal.azure.com/#blade/HubsExtension/BrowseResourceBlade/resourceType/Microsoft.Search%2FsearchServices) under your current subscription. You can use a free service for this quickstart.

-+ An Azure Storage account with [Blob storage](../storage/blobs/index.yml), used for hosting sample data, and for persisting temporary data created during a debug session.

-+ [Sample data (clinical trials)](https://github.com/Azure-Samples/azure-search-sample-data/tree/master/clinical-trials/clinical-trials-pdf-19).

-> This quickstart also uses [Azure Cognitive Services](https://azure.microsoft.com/services/cognitive-services/) for the AI. Because the workload is so small, Cognitive Services is tapped behind the scenes for free processing for up to 20 transactions. This means that you can complete this exercise without having to create an additional Cognitive Services resource.

-1. Start Postman and import the collection. Under **Files** > **New**, select the collection to import.

-1. Click **Edit**.

-1. Under Current Value, enter the name of your `searchService` (for example, if the endpoint is `https://mydemo.search.windows.net`, then the service name is "`mydemo`").

-1. Enter the `apiKey` with either the primary or secondary key of your search service.

-1. Enter the `storageConnectionString` from the keys page of your Azure Storage account.

-1. Enter the `containerName` for the container you created in the storage account, and then click **Update**.

- :::image type="content" source="media/cognitive-search-debug/postman-enter-variables.png" alt-text="edit variables in Postman":::

-The collection contains four different REST calls that are used to create objects used in this tutorial.

-The first call creates the data source. `clinical-trials-ds`. The second call creates the skillset, `clinical-trials-ss`. The third call creates the index, `clinical-trials`. The fourth and final call creates the indexer, `clinical-trials-idxr`.

-+ Open each request in turn, and click **Send** to send each request to the search service.

-After all of the calls in the collection have been completed, close Postman and return to the Azure portal.

-The sample code intentionally creates a buggy index as a consequence of problems that occurred during skillset execution. The problem is missing data.

-1. In Azure portal, on the search service Overview page, select the **Indexes** tab.

-1. Select the `clinical-trials` index.

-1. Click **Search** to run the query, returning all 19 documents, showing empty values for "organizations" and "locations".

-These fields should have been populated through the skillset's [Entity Recognition skill](cognitive-search-skill-entity-recognition-v3.md), used to find organizations and locations anywhere within the blob's content. In the next exercise, you'll use Debug session to determine what went wrong.

-1. Open the **Indexers** tab and select `clinical-trials-idxr`.

-1. Notice that while the indexer job succeeded overall, there were 57 warnings.

-1. Click **Success** to view the warnings (if there were mostly errors, the detail link would be **Failed**). You'll see a long list of every warning emitted by the indexer.

- :::image type="content" source="media/cognitive-search-debug/portal-indexer-errors-warnings.png" alt-text="view warnings":::

-1. From the search overview page, click the **Debug sessions** tab.

-1. Accept the default document choice for the first document in the collection.

-1. **Save** the session. Saving the session will kick off the AI enrichment pipeline as defined by the skillset.

-> [!Important]

-> A debug session only works with a single document. You can choose which document to debug, or just use the first one.

-<!-- > :::image type="content" source="media/cognitive-search-debug/debug-execution-complete1.png" alt-text="New debug session started"::: -->

-When the debug session has finished initializing, the session defaults to the **AI Enrichments** tab, highlighting the **Skill Graph**. The Skill Graph provides a visual hierarchy of the skillset and its order of execution sequentially and in parallel.

-Any issues reported by the indexer can be found in the adjacent **Errors/Warnings** tab.

-Select **Errors/Warnings** to review the notifications. You should see three:

- + "Could not map output field 'locations' to search index. Check the 'outputFieldMappings' property of your indexer.

- + "Could not map output field 'organizations' to search index. Check the 'outputFieldMappings' property of your indexer.

- + "Skill executed but may have unexpected results because one or more skill inputs was invalid.

- Many skills have a 'languageCode' parameter. By inspecting the operation, you can see that this language code input is missing from the `Enrichment.NerSkillV2.#1`, which is the same Entity Recognition skill that is having trouble with 'locations' and 'organizations' output.

-Because all three notifications are about this skill, your next step is to debug this skill. If possible, start by solving input issues first before moving on to outputFieldMapping issues.

- :::image type="content" source="media/cognitive-search-debug/debug-session-errors-warnings.png" alt-text="New debug session started":::

-<!-- + The Skill Graph provides a visual hierarchy of the skillset and its order of execution from top to bottom. Skills that are side by side in the graph are executed in parallel. Color coding of skills in the graph indicate the types of skills that are being executed in the skillset. In the example, the green skills are text and the red skill is vision. Clicking on an individual skill in the graph will display the details of that instance of the skill in the right pane of the session window. The skill settings, a JSON editor, execution details, and errors/warnings are all available for review and editing.

-+ The Enriched Data Structure details the nodes in the enrichment tree generated by the skills from the source document's contents. -->

-## Fix missing skill input value

-In the **Errors/Warnings** tab, there is an error for an operation labeled `Enrichment.NerSkillV2.#1`. The detail of this error explains that there is a problem with a skill input value '/document/languageCode'.

-1. Return to the **AI Enrichments** tab.

-1. Click on the **Skill Graph**.

-1. Click on the skill labeled **#1** to display its details in the right pane.

-1. Locate the input for "languageCode".

-1. Select the **</>** symbol at the beginning of the line and open the Expression Evaluator.

-1. Click the **Evaluate** button to confirm that this expression is resulting in an error. It will confirm that the "languageCode" property is not a valid input.

- :::image type="content" source="media/cognitive-search-debug/expression-evaluator-language.png" alt-text="Expression Evaluator":::

-There are two ways to research this error in the session. The first is to look at where the input is coming from - what skill in the hierarchy is supposed to produce this result? The Executions tab in the skill details pane should display the source of the input. If there is no source, this indicates a field mapping error.

-1. Click the **Executions** tab.

-1. Look at the INPUTS and find "languageCode". There is no source for this input listed.

-1. Switch the left pane to display the Enriched Data Structure. There is no mapped path corresponding to "languageCode".

- :::image type="content" source="media/cognitive-search-debug/enriched-data-structure-language.png" alt-text="Enriched Data Structure":::

-There is a mapped path for "language." So, there is a typo in the skill settings. To fix this expression in the #1 skill with the expression '/document/language' will need to be updated.

-1. Open the Expression Evaluator **</>** for the path "language."

-1. Copy the expression. Close the window.

-1. Go to the Skill Settings for the #1 skill and open the Expression Evaluator **</>** for the input "languageCode."

-1. Paste the new value, '/document/language' into the Expression box and click **Evaluate**.

-1. It should display the correct input "en". Click Apply to update the expression.

-1. Click **Save** in the right, skill details pane.

-1. Click **Run** in the session's window menu. This will kick off another execution of the skillset using the document.

-Once the debug session execution completes, click the Errors/Warnings tab and it will show that the error labeled "Enrichment.NerSkillV2.#1" is gone. However, there are still two warnings that the service could not map output fields for organizations and locations to the search index. There are missing values: '/document/merged_content/organizations' and '/document/merged_content/locations'.

-## Fix missing skill output values

-There are missing output values from a skill. To identify the skill with the error go to the Enriched Data Structure, find the value name and look at its Originating Source. In the case of the missing organizations and locations values, they are outputs from skill #1. Opening the Expression Evaluator </> for each path, will display the expressions listed as '/document/content/organizations' and '/document/content/locations', respectively.

-The output for these entities is empty and it should not be empty. What are the inputs producing this result?

-1. Go to **Skill Graph** and select skill #1.

-1. Select **Executions** tab in the right skill details pane.

-1. Open the Expression Evaluator **</>** for the INPUT "text."

- :::image type="content" source="media/cognitive-search-debug/input-skill-missing-value-locations-organizations.png" alt-text="Input for text skill":::

-The displayed result for this input doesnΓÇÖt look like a text input. It looks like an image that is surrounded by new lines. The lack of text means that no entities can be identified. Looking at the hierarchy of the skillset displays the content is first processed by the #6 (OCR) skill and then passed to the #5 (Merge) skill.

-1. Select the #5 (Merge) skill in the **Skill Graph**.

-1. Select the **Executions** tab in the right skill details pane and open the Expression Evaluator **</>** for the OUTPUTS "mergedText".

- :::image type="content" source="media/cognitive-search-debug/merge-output-detail-missing-value-locations-organizations.png" alt-text="Output for Merge skill":::

-Here the text is paired with the image. Looking at the expression '/document/merged_content' the error in the "organizations" and "locations" paths for the #1 skill is visible. Instead of using '/document/content' it should use '/document/merged_content' for the "text" inputs.

-1. Copy the expression for the "mergedText" output and close the Expression Evaluator window.

-1. Select skill #1 in the **Skill Graph**.

-1. Select the **Skill Settings** tab in the right skill details pane.

-1. Open the Expression Evaluator **</>** for the "text" input.

-1. Paste the new expression into the box. Click **Evaluate**.

-1. The correct input with the added text should be displayed. Click **Apply** to update the Skill Settings.

-1. Click **Save** in the right, skill details pane.

-1. Click **Run** in the sessions window menu. This will kick off another execution of the skillset using the document.

-After the indexer has finished running, the errors are still there. Go back to skill #1 and investigate. The input to the skill was corrected to 'merged_content' from 'content'. What are the outputs for these entities in the skill?

-1. Select the **AI Enrichments** tab.

-1. Select **Skill Graph** and click on skill #1.

-1. Navigate the **Skill Settings** to find "outputs."

-1. Open the Expression Evaluator **</>** for the "organizations" entity.

- :::image type="content" source="media/cognitive-search-debug/skill-output-detail-missing-value-locations-organizations.png" alt-text="Output for organizations entity":::

-Evaluating the result of the expression gives the correct result. The skill is working to identify the correct value for the entity, "organizations." However, the output mapping in the entity's path is still throwing an error. In comparing the output path in the skill to the output path in the error, the skill that is parenting the outputs, organizations and locations under the /document/content node. While the output field mapping is expecting the results to be parented under the /document/merged_content node. In the previous step, the input changed from '/document/content' to '/document/merged_content'. The context in the skill settings needs to be changed in order to ensure the output is generated with the right context.

-1. Select the **AI Enrichments** tab.

-1. Select **Skill Graph** and click on skill #1.

-1. Click **Save** in the right, skill details pane.

-1. Click **Run** in the sessions window menu. This will kick off another execution of the skillset using the document.

- :::image type="content" source="media/cognitive-search-debug/skill-setting-context-correction-missing-value-locations-organizations.png" alt-text="Context correction in skill setting":::

-1. Click **Commit changes** in the main Debug sessions menu.

-1. Click **OK** to confirm that you wish to update your skillset.

-1. Click **Reset**.

-1. Click **Run**. Click **OK** to confirm.

-1. Click **Search**.

-In this article, learn how to request a semantic answer, unpack the response, and what content characteristics are most conducive to producing high quality answers.

-## How to specify "answers" in a query request

-The approach for listing fields in priority order has changed recently, with semanticConfiguration replacing searchFields. If you are currently using searchFields, please update your code to the 2021-04-30-Preview API version and use semanticConfiguration instead.

-The "semanticConfiguration" parameter is crucial to returning a high quality answer.

-The "searchFields" parameter is crucial to returning a high quality answer, both in terms of content and order (see below).

-## Deconstruct an answer from the response

-Answers are provided in the @search.answers array, which appears first in the response. If an answer is indeterminate, the response will show up as `"@search.answers": []`. When designing a search results page that includes answers, be sure to handle cases where answers are not found.

-Within @search.answers, the "key" is the document key or ID of the match. Given a document key, you can use [Lookup Document](/rest/api/searchservice/lookup-document) API to retrieve any or all parts of the search document to include on the search page or a detail page.

-Both "text" and "highlights" provide identical content, in both plain text and with highlights. By default, highlights are styled as `<em>`, which you can override using the existing highlightPreTag and highlightPostTag parameters. As noted elsewhere, the substance of an answer is verbatim content from a search document. The extraction model looks for characteristics of an answer to find the appropriate content, but does not compose new language in the response.

-The "score" is a confidence score that reflects the strength of the answer. If there are multiple answers in the response, this score is used to determine the order. Top answers and top captions can be derived from different search documents, where the top answer originates from one document, and the top caption from another, but in general you will see the same documents in the top positions within each array.

-Answers are followed by the "value" array, which always includes scores, captions, and any fields that are retrievable by default. If you specified the select parameter, the "value" array is limited to the fields that you specified. For more information about items in the response, see [Create a semantic query](semantic-how-to-query-request.md).

-Given the query "how do clouds form", the following answer is returned in the response:

-+ query strings must not be null (search=`*`) and the string should have the characteristics of a question, as opposed to a keyword search (a sequential list of arbitrary terms or phrases). If the query string does not appear to be answer, answer processing is skipped, even if the request specifies "answers" as a query parameter.

-Consumption models in Synapse SQL enables you to use different database objects. The comparison of supported object types is shown in the following table:

-| **INSERT statement** | Yes | No |

-| **UPDATE statement** | Yes | No |

-| **DELETE statement** | Yes | No |

-| **MERGE statement** | Yes ([preview](/sql/t-sql/statements/merge-transact-sql?view=azure-sqldw-latest&preserve-view=true)) | No |

-| **Data load** | Yes. Preferred utility is [COPY](/sql/t-sql/statements/copy-into-transact-sql?view=azure-sqldw-latest&preserve-view=true) statement, but the system supports both BULK load (BCP) and [CETAS](/sql/t-sql/statements/create-external-table-as-select-transact-sql?view=azure-sqldw-latest&preserve-view=true) for data loading. | No |

-| **SERVER SCOPED CREDENTIAL** | No | Yes |

-| **Permissions - [Server-level](/sql/relational-databases/security/authentication-access/server-level-roles)** | No | Yes |

-| **DATABASE SCOPED CREDENTIAL** | Yes | Yes |

-| **Row-level security** | [Yes](/sql/relational-databases/security/row-level-security?toc=/azure/synapse-analytics/sql-data-warehouse/toc.json&bc=/azure/synapse-analytics/sql-data-warehouse/breadcrumb/toc.json&view=azure-sqldw-latest&preserve-view=true) | No |

-| **Auditing** | [Yes](../../azure-sql/database/auditing-overview.md) | No |

-| **Internal storage** | Yes | No |

-| **Azure Data Lake v2** | Yes | Yes |

-| **Azure Blob Storage** | Yes | Yes |

-| **Azure SQL (remote)** | No | No |

-| **Azure CosmosDB transactional storage** | No | No |

-| **Apache Spark tables (in workspace)** | No | PARQUET tables only using [metadata synchronization](develop-storage-files-spark-tables.md) |

-| **[Delta-lake](https://delta.io/)** | No | [Yes](query-delta-lake-format.md) |

-Wrong network configuration is often the cause for this behaviour. Make sure the ports are appropriatelly configured. In case you use firewall or Private Endpoint check their settings as well. Finally, make sure the appropriate roles are granted.