Azure Monitor Query client library for Python

The Azure Monitor Query client library is used to execute read-only queries against Azure Monitor's two data platforms:

• Logs - Collects and organizes log and performance data from monitored resources. Data from different sources such as platform logs from Azure services, log and performance data from virtual machines agents, and usage and performance data from apps can be consolidated into a single Azure Log Analytics workspace. The various data types can be analyzed together using the Kusto Query Language.
• Metrics - Collects numeric data from monitored resources into a time series database. Metrics are numerical values that are collected at regular intervals and describe some aspect of a system at a particular time. Metrics are lightweight and capable of supporting near real-time scenarios, making them useful for alerting and fast detection of issues.

Resources:

• Source code
• Package (PyPI)
• Package (Conda)
• API reference documentation
• Service documentation
• Samples
• Change log

Getting started

Prerequisites

• Python 3.7 or later
• An Azure subscription
• A TokenCredential implementation, such as an Azure Identity library credential type.
• To query Logs, you need one of the following things:
  • An Azure Log Analytics workspace
  • An Azure resource of any kind (Storage Account, Key Vault, Cosmos DB, etc.)
• To query Metrics, you need an Azure resource of any kind (Storage Account, Key Vault, Cosmos DB, etc.).

Install the package

Install the Azure Monitor Query client library for Python with pip:

pip install azure-monitor-query

Create the client

An authenticated client is required to query Logs or Metrics. The library includes both synchronous and asynchronous forms of the clients. To authenticate, create an instance of a token credential. Use that instance when creating a LogsQueryClient, MetricsQueryClient, or MetricsBatchQueryClient. The following examples use DefaultAzureCredential from the azure-identity package.

Synchronous clients

Consider the following example, which creates synchronous clients for both Logs and Metrics querying:

from azure.identity import DefaultAzureCredential
from azure.monitor.query import LogsQueryClient, MetricsQueryClient

credential = DefaultAzureCredential()
logs_client = LogsQueryClient(credential)
metrics_client = MetricsQueryClient(credential)

Asynchronous clients

The asynchronous forms of the query client APIs are found in the .aio-suffixed namespace. For example:

from azure.identity.aio import DefaultAzureCredential
from azure.monitor.query.aio import LogsQueryClient, MetricsQueryClient

credential = DefaultAzureCredential()
async_logs_client = LogsQueryClient(credential)
async_metrics_client = MetricsQueryClient(credential)

Configure clients for non-public Azure clouds

By default, LogsQueryClient and MetricsQueryClient are configured to connect to the public Azure cloud. These can be configured to connect to non-public Azure clouds by passing in the correct endpoint argument: For example:

logs_client = LogsQueryClient(credential, endpoint="https://api.loganalytics.azure.cn/v1")
metrics_client = MetricsQueryClient(credential, endpoint="https://management.chinacloudapi.cn")

Note: Currently, MetricsQueryClient uses the Azure Resource Manager (ARM) endpoint for querying metrics, so you will need the corresponding management endpoint for your cloud when using this client. This is subject to change in the future.

Execute the query

For examples of Logs and Metrics queries, see the Examples section.

Key concepts

Logs query rate limits and throttling

The Log Analytics service applies throttling when the request rate is too high. Limits, such as the maximum number of rows returned, are also applied on the Kusto queries. For more information, see Query API.

If you're executing a batch logs query, a throttled request will return a LogsQueryError object. That object's code value will be ThrottledError.

Metrics data structure

Each set of metric values is a time series with the following characteristics:

• The time the value was collected
• The resource associated with the value
• A namespace that acts like a category for the metric
• A metric name
• The value itself
• Some metrics may have multiple dimensions as described in multi-dimensional metrics. Custom metrics can have up to 10 dimensions.

Examples

• Logs query
  • Specify timespan
  • Handle logs query response
• Batch logs query
• Resource logs query
• Advanced logs query scenarios
  • Set logs query timeout
  • Query multiple workspaces
  • Include statistics
  • Include visualization
• Metrics query
  • Handle metrics query response
  • Example of handling response

Logs query

This example shows how to query a Log Analytics workspace. To handle the response and view it in a tabular form, the pandas library is used. See the samples if you choose not to use pandas.

Specify timespan

The timespan parameter specifies the time duration for which to query the data. This value can be one of the following:

• a timedelta
• a timedelta and a start datetime
• a start datetime/end datetime

For example:

import os
import pandas as pd
from datetime import datetime, timezone
from azure.monitor.query import LogsQueryClient, LogsQueryStatus
from azure.identity import DefaultAzureCredential
from azure.core.exceptions import HttpResponseError

credential = DefaultAzureCredential()
client = LogsQueryClient(credential)

query = """AppRequests | take 5"""

start_time=datetime(2021, 7, 2, tzinfo=timezone.utc)
end_time=datetime(2021, 7, 4, tzinfo=timezone.utc)

try:
    response = client.query_workspace(
        workspace_id=os.environ['LOG_WORKSPACE_ID'],
        query=query,
        timespan=(start_time, end_time)
        )
    if response.status == LogsQueryStatus.PARTIAL:
        error = response.partial_error
        data = response.partial_data
        print(error)
    elif response.status == LogsQueryStatus.SUCCESS:
        data = response.tables
    for table in data:
        df = pd.DataFrame(data=table.rows, columns=table.columns)
        print(df)
except HttpResponseError as err:
    print("something fatal happened")
    print(err)

Handle logs query response

The query_workspace API returns either a LogsQueryResult or a LogsQueryPartialResult object. The batch_query API returns a list that may contain LogsQueryResult, LogsQueryPartialResult, and LogsQueryError objects. Here's a hierarchy of the response:

LogsQueryResult
|---statistics
|---visualization
|---tables (list of `LogsTable` objects)
    |---name
    |---rows
    |---columns
    |---columns_types

LogsQueryPartialResult
|---statistics
|---visualization
|---partial_error (a `LogsQueryError` object)
    |---code
    |---message
    |---details
    |---status
|---partial_data (list of `LogsTable` objects)
    |---name
    |---rows
    |---columns
    |---columns_types

The LogsQueryResult directly iterates over the table as a convenience. For example, to handle a logs query response with tables and display it using pandas:

response = client.query(...)
for table in response:
    df = pd.DataFrame(table.rows, columns=[col.name for col in table.columns])

A full sample can be found here.

In a similar fashion, to handle a batch logs query response:

for result in response:
    if result.status == LogsQueryStatus.SUCCESS:
        for table in result:
            df = pd.DataFrame(table.rows, columns=table.columns)
            print(df)

A full sample can be found here.

Batch logs query

The following example demonstrates sending multiple queries at the same time using the batch query API. The queries can either be represented as a list of LogsBatchQuery objects or a dictionary. This example uses the former approach.

import os
from datetime import timedelta, datetime, timezone
import pandas as pd
from azure.monitor.query import LogsQueryClient, LogsBatchQuery, LogsQueryStatus
from azure.identity import DefaultAzureCredential

credential = DefaultAzureCredential()
client = LogsQueryClient(credential)
requests = [
    LogsBatchQuery(
        query="AzureActivity | summarize count()",
        timespan=timedelta(hours=1),
        workspace_id=os.environ['LOG_WORKSPACE_ID']
    ),
    LogsBatchQuery(
        query= """bad query""",
        timespan=timedelta(days=1),
        workspace_id=os.environ['LOG_WORKSPACE_ID']
    ),
    LogsBatchQuery(
        query= """let Weight = 92233720368547758;
        range x from 1 to 3 step 1
        | summarize percentilesw(x, Weight * 100, 50)""",
        workspace_id=os.environ['LOG_WORKSPACE_ID'],
        timespan=(datetime(2021, 6, 2, tzinfo=timezone.utc), datetime(2021, 6, 5, tzinfo=timezone.utc)), # (start, end)
        include_statistics=True
    ),
]
results = client.query_batch(requests)

for res in results:
    if res.status == LogsQueryStatus.FAILURE:
        # this will be a LogsQueryError
        print(res.message)
    elif res.status == LogsQueryStatus.PARTIAL:
        ## this will be a LogsQueryPartialResult
        print(res.partial_error)
        for table in res.partial_data:
            df = pd.DataFrame(table.rows, columns=table.columns)
            print(df)
    elif res.status == LogsQueryStatus.SUCCESS:
        ## this will be a LogsQueryResult
        table = res.tables[0]
        df = pd.DataFrame(table.rows, columns=table.columns)
        print(df)

Resource logs query

The following example demonstrates how to query logs directly from an Azure resource without the use of a Log Analytics workspace. Here, the query_resource method is used instead of query_workspace, and instead of a workspace ID, an Azure resource identifier is passed in (e.g. /subscriptions/{subscription-id}/resourceGroups/{resource-group-name}/providers/{resource-provider}/{resource-type}/{resource-name}).

import os
import pandas as pd
from datetime import timedelta
from azure.monitor.query import LogsQueryClient, LogsQueryStatus
from azure.core.exceptions import HttpResponseError
from azure.identity import DefaultAzureCredential

credential  = DefaultAzureCredential()
client = LogsQueryClient(credential)

query = """AzureActivity | take 5"""

try:
    response = client.query_resource(os.environ['LOGS_RESOURCE_ID'], query, timespan=timedelta(days=1))
    if response.status == LogsQueryStatus.PARTIAL:
        error = response.partial_error
        data = response.partial_data
        print(error)
    elif response.status == LogsQueryStatus.SUCCESS:
        data = response.tables
    for table in data:
        df = pd.DataFrame(data=table.rows, columns=table.columns)
        print(df)
except HttpResponseError as err:
    print("something fatal happened")
    print(err)

Advanced logs query scenarios

Set logs query timeout

The following example shows setting a server timeout in seconds. A gateway timeout is raised if the query takes more time than the mentioned timeout. The default is 180 seconds and can be set up to 10 minutes (600 seconds).

import os
from datetime import timedelta
from azure.monitor.query import LogsQueryClient
from azure.identity import DefaultAzureCredential

credential = DefaultAzureCredential()
client = LogsQueryClient(credential)

response = client.query_workspace(
    os.environ['LOG_WORKSPACE_ID'],
    "range x from 1 to 10000000000 step 1 | count",
    timespan=timedelta(days=1),
    server_timeout=600 # sets the timeout to 10 minutes
    )

Query multiple workspaces

The same logs query can be executed across multiple Log Analytics workspaces. In addition to the Kusto query, the following parameters are required:

• workspace_id - The first (primary) workspace ID.
• additional_workspaces - A list of workspaces, excluding the workspace provided in the workspace_id parameter. The parameter's list items may consist of the following identifier formats:
  • Qualified workspace names
  • Workspace IDs
  • Azure resource IDs

For example, the following query executes in three workspaces:

client.query_workspace(
    <workspace_id>,
    query,
    timespan=timedelta(days=1),
    additional_workspaces=['<workspace 2>', '<workspace 3>']
    )

A full sample can be found here.

Include statistics

To get logs query execution statistics, such as CPU and memory consumption:

1. Set the include_statistics parameter to True.
2. Access the statistics field inside the LogsQueryResult object.

The following example prints the query execution time:

query = "AzureActivity | top 10 by TimeGenerated"
result = client.query_workspace(
    <workspace_id>,
    query,
    timespan=timedelta(days=1),
    include_statistics=True
    )

execution_time = result.statistics.get("query", {}).get("executionTime")
print(f"Query execution time: {execution_time}")

The statistics field is a dict that corresponds to the raw JSON response, and its structure can vary by query. The statistics are found within the query property. For example:

{
  "query": {
    "executionTime": 0.0156478,
    "resourceUsage": {...},
    "inputDatasetStatistics": {...},
    "datasetStatistics": [{...}]
  }
}

Include visualization

To get visualization data for logs queries using the render operator:

1. Set the include_visualization property to True.
2. Access the visualization field inside the LogsQueryResult object.

For example:

query = (
    "StormEvents"
    "| summarize event_count = count() by State"
    "| where event_count > 10"
    "| project State, event_count"
    "| render columnchart"
)
result = client.query_workspace(
    <workspace_id>,
    query,
    timespan=timedelta(days=1),
    include_visualization=True
    )

print(f"Visualization result: {result.visualization}")

The visualization field is a dict that corresponds to the raw JSON response, and its structure can vary by query. For example:

{
  "visualization": "columnchart",
  "title": "the chart title",
  "accumulate": False,
  "isQuerySorted": False,
  "kind": None,
  "legend": None,
  "series": None,
  "yMin": "NaN",
  "yMax": "NaN",
  "xAxis": None,
  "xColumn": None,
  "xTitle": "x axis title",
  "yAxis": None,
  "yColumns": None,
  "ySplit": None,
  "yTitle": None,
  "anomalyColumns": None
}

Interpretation of the visualization data is left to the library consumer. To use this data with the Plotly graphing library, see the synchronous or asynchronous code samples.

Metrics query

The following example gets metrics for an Event Grid subscription. The resource URI is that of an Event Grid topic.

The resource URI must be that of the resource for which metrics are being queried. It's normally of the format /subscriptions/<id>/resourceGroups/<rg-name>/providers/<source>/topics/<resource-name>.

To find the resource URI:

1. Navigate to your resource's page in the Azure portal.
2. From the Overview blade, select the JSON View link.
3. In the resulting JSON, copy the value of the id property.

NOTE: The metrics are returned in the order of the metric_names sent.

import os
from datetime import timedelta, datetime
from azure.monitor.query import MetricsQueryClient
from azure.identity import DefaultAzureCredential

credential = DefaultAzureCredential()
client = MetricsQueryClient(credential)
start_time = datetime(2021, 5, 25)
duration = timedelta(days=1)
metrics_uri = os.environ['METRICS_RESOURCE_URI']
response = client.query_resource(
    metrics_uri,
    metric_names=["PublishSuccessCount"],
    timespan=(start_time, duration)
    )

for metric in response.metrics:
    print(metric.name)
    for time_series_element in metric.timeseries:
        for metric_value in time_series_element.data:
            print(metric_value.time_stamp)

Handle metrics query response

The metrics query API returns a MetricsQueryResult object. The MetricsQueryResult object contains properties such as a list of Metric-typed objects, granularity, namespace, and timespan. The Metric objects list can be accessed using the metrics param. Each Metric object in this list contains a list of TimeSeriesElement objects. Each TimeSeriesElement object contains data and metadata_values properties. In visual form, the object hierarchy of the response resembles the following structure:

MetricsQueryResult
|---granularity
|---timespan
|---cost
|---namespace
|---resource_region
|---metrics (list of `Metric` objects)
    |---id
    |---type
    |---name
    |---unit
    |---timeseries (list of `TimeSeriesElement` objects)
        |---metadata_values
        |---data (list of data points represented by `MetricValue` objects)

Example of handling response

import os
from azure.monitor.query import MetricsQueryClient, MetricAggregationType
from azure.identity import DefaultAzureCredential

credential = DefaultAzureCredential()
client = MetricsQueryClient(credential)

metrics_uri = os.environ['METRICS_RESOURCE_URI']
response = client.query_resource(
    metrics_uri,
    metric_names=["MatchedEventCount"],
    aggregations=[MetricAggregationType.COUNT]
    )

for metric in response.metrics:
    print(metric.name)
    for time_series_element in metric.timeseries:
        for metric_value in time_series_element.data:
            if metric_value.count != 0:
                print(
                    "There are {} matched events at {}".format(
                        metric_value.count,
                        metric_value.time_stamp
                    )
                )

Metrics batch query

A user can also query metrics from multiple resources at once using the query_batch method of MetricsBatchQueryClient. This uses a different API than the MetricsQueryClient and requires that a user pass in a regional endpoint when instantiating the client (for example, "https://westus3.metrics.monitor.azure.com").

Note, each resource must be in the same region as the endpoint passed in when instantiating the client, and each resource must be in the same Azure subscription. Furthermore, the metric namespace that contains the metrics to be queried must also be passed. A list of metric namespaces can be found here.

from datetime import timedelta
import os

from azure.core.exceptions import HttpResponseError
from azure.identity import DefaultAzureCredential
from azure.monitor.query import MetricsBatchQueryClient, MetricAggregationType


credential = DefaultAzureCredential()
client = MetricsBatchQueryClient(endpoint, credential)

resource_uris = [
    "/subscriptions/<id>/resourceGroups/<rg-name>/providers/<source>/storageAccounts/<resource-name-1>",
    "/subscriptions/<id>/resourceGroups/<rg-name>/providers/<source>/storageAccounts/<resource-name-2>"
]

response = client.query_batch(
    resource_uris,
    metric_namespace="Microsoft.Storage/storageAccounts",
    metric_names=["Ingress"],
    timespan=timedelta(hours=2),
    granularity=timedelta(minutes=5),
    aggregations=[MetricAggregationType.AVERAGE],
)

for metrics_query_result in response:
    print(metrics_query_result.timespan)

Troubleshooting

See our troubleshooting guide for details on how to diagnose various failure scenarios.

Next steps

To learn more about Azure Monitor, see the Azure Monitor service documentation.

Samples

The following code samples show common scenarios with the Azure Monitor Query client library.

Logs query samples

• Send a single query with LogsQueryClient and handle the response as a table (async sample)
• Send a single query with LogsQueryClient and handle the response in key-value form
• Send a single query with LogsQueryClient without pandas
• Send a single query with LogsQueryClient across multiple workspaces
• Send multiple queries with LogsQueryClient
• Send a single query with LogsQueryClient using server timeout

Metrics query samples

• Send a query using MetricsQueryClient (async sample)
• Get a list of metric namespaces (async sample)
• Get a list of metric definitions (async sample)

Contributing

This project welcomes contributions and suggestions. Most contributions require you to agree to a Contributor License Agreement (CLA) declaring that you have the right to, and actually do, grant us the rights to use your contribution. For details, visit cla.microsoft.com.

When you submit a pull request, a CLA-bot will automatically determine whether you need to provide a CLA and decorate the PR appropriately (e.g., label, comment). Simply follow the instructions provided by the bot. You will only need to do this once across all repositories using our CLA.

This project has adopted the Microsoft Open Source Code of Conduct. For more information see the Code of Conduct FAQ or contact opencode@microsoft.com with any additional questions or comments.

Release History

1.3.0b2 (2023-11-20)

Other Changes

• Internal updates to generated code.
• Bumped minimum dependency on azure-core to >=1.28.0.

1.3.0b1 (2023-08-16)

Features Added

• Added MetricsBatchQueryClient to support batch querying metrics from Azure resources. (#31049)

1.2.0 (2023-05-09)

Features Added

• Add the query_resource method to LogsQueryClient to allow users to query Azure resources directly without the context of a workspace. (#29365)

Bugs Fixed

• Fixed an inconsistent keyword argument name in the LogsTable constructor, changing column_types to columns_types. Note that this is a class that is typically only instantiated internally, and not by users. (#29076)

Other Changes

• Improved client configuration logic for non-public Azure clouds where credential scope will be determined based on the configured endpoint. (#29602)

1.1.1 (2023-02-13)

Bugs Fixed

• Fixed a bug where the incorrect key time_stamp (should be timeStamp) was used in the creation of MetricValue objects (thanks @jamespic). (#28777)

1.1.0 (2023-02-07)

Bugs Fixed

• Error details are now propagated inside the LogsQueryError object. (#25137)

Other Changes

• Python 3.6 is no longer supported. Please use Python version 3.7 or later. For more details, see Azure SDK for Python version support policy.
• Removed msrest dependency.
• Bumped minimum dependency on azure-core to >=1.24.0.
• Added requirement for isodate>=0.6.0 (isodate was required by msrest).
• Added requirement for typing-extensions>=4.0.1.

1.0.3 (2022-07-07)

Bugs Fixed

• Fixed a bug where query_resource in metrics client is throwing an error with unexpected metric_namespace argument.

1.0.2 (2022-05-06)

• This version and all future versions will require Python 3.6+. Python 2.7 is no longer supported.

Bugs Fixed

• Fixed a bug where having a None value in datetime throws

1.0.1 (2021-11-09)

Bugs Fixed

• Fixed a bug where Metadata values in timestamp don't show up sometimes.

1.0.0 (2021-10-06)

Features Added

• Added LogsQueryPartialResult and LogsQueryError to handle errors.
• Added status attribute to LogsQueryResult.
• Added LogsQueryStatus Enum to describe the status of a result.
• Added a new LogsTableRow type that represents a single row in a table.
• Items in metrics list in MetricsQueryResult can now be accessed by metric names.

Breaking Changes

• LogsQueryResult now iterates over the tables directly as a convenience.
• query API in logs is renamed to query_workspace
• query API in metrics is renamed to query_resource
• query_workspace API now returns a union of LogsQueryPartialResult and LogsQueryResult.
• query_batch API now returns a union of LogsQueryPartialResult, LogsQueryError and LogsQueryResult.
• metric_namespace is renamed to namespace and is a keyword-only argument in list_metric_definitions API.
• MetricsResult is renamed to MetricsQueryResult.

1.0.0b4 (2021-09-09)

Features Added

• Added additional display_description attribute to the Metric type.
• Added a MetricClass enum to provide the class of a metric.
• Added a metric_class attribute to the MetricDefinition type.
• Added a MetricNamespaceClassification enum to support the namespace_classification attribute on MetricNamespace type.
• Added a MetricUnit enum to describe the unit of the metric.

Breaking Changes

• Rename batch_query to query_batch.
• Rename LogsBatchQueryRequest to LogsBatchQuery.
• include_render is now renamed to include_visualization in the query API.
• LogsQueryResult now returns visualization instead of render.
• start_time, duration and end_time are now replaced with a single param called timespan
• resourceregion is renamed to resource_region in the MetricResult type.
• top is renamed to max_results in the metric's query API.
• metric_namespace_name is renamed to fully_qualified_namespace
• is_dimension_required is renamed to dimension_required
• interval and time_grain are renamed to granularity
• orderby is renamed to order_by
• LogsQueryResult now returns datetime objects for a time values.
• LogsBatchQuery doesn't accept a request_id anymore.
• MetricsMetadataValues is removed. A dictionary is used instead.
• time_stamp is renamed to timestamp in MetricValue type.
• AggregationType is renamed to MetricAggregationType.
• Removed LogsBatchResultError type.
• LogsQueryResultTable is named to LogsTable
• LogsTableColumn is now removed. Column labels are strings instead.
• start_time in list_metric_namespaces API is now a datetime.
• The order of params in LogsBatchQuery is changed. Also, headers is no longer accepted.
• timespan is now a required keyword-only argument in logs APIs.
• batch api now returns a list of LogsQueryResult objects.

Bugs Fixed

• include_statistics and include_visualization args can now work together.

1.0.0b3 (2021-08-09)

Features Added

• Added enum AggregationType which can be used to specify aggregations in the query API.
• Added LogsBatchQueryResult model that is returned for a logs batch query.
• Added error attribute to LogsQueryResult.

Breaking Changes

• aggregation param in the query API is renamed to aggregations
• batch_query API now returns a list of responses.
• LogsBatchResults model is now removed.
• LogsQueryRequest is renamed to LogsBatchQueryRequest
• LogsQueryResults is now renamed to LogsQueryResult
• LogsBatchQueryResult now has 4 additional attributes - tables, error, statistics and render instead of body attribute.

1.0.0b2 (2021-07-06)

Breaking Changes

• workspaces, workspace_ids, qualified_names and azure_resource_ids are now merged into a single additional_workspaces list in the query API.
• The LogQueryRequest object now takes in a workspace_id and additional_workspaces instead of workspace.
• aggregation param is now a list instead of a string in the query method.
• duration must now be provided as a timedelta instead of a string.

1.0.0b1 (2021-06-10)

Features

• Version (1.0.0b1) is the first preview of our efforts to create a user-friendly and Pythonic client library for Azure Monitor Query. For more information about this, and preview releases of other Azure SDK libraries, please visit https://azure.github.io/azure-sdk/releases/latest/python.html.
• Added ~azure.monitor.query.LogsQueryClient to query log analytics along with ~azure.monitor.query.aio.LogsQueryClient.
• Implements the ~azure.monitor.query.MetricsQueryClient for querying metrics, listing namespaces and metric definitions along with ~azure.monitor.query.aio.MetricsQueryClient.