Azure Monitor

For general information on resource management, see Resource Management.

Create the Monitor client

The following code creates an instance of the client.

You will need to provide your subscription_id which can be retrieved from your subscription list.

See Resource Management Authentication for details on handling Azure Active Directory authentication with the Python SDK, and creating a Credentials instance.

from azure.monitor import MonitorClient
from azure.mgmt.monitor import MonitorMgmtClient
from azure.common.credentials import UserPassCredentials

# Replace this with your subscription id
subscription_id = '33333333-3333-3333-3333-333333333333'

# See above for details on creating different types of AAD credentials
credentials = UserPassCredentials(
    'user@domain.com',  # Your user
    'my_password',      # Your password
)

client = MonitorClient(
    credentials,
    subscription_id
)

monitor_mgmt_client = MonitorMgmtClient(
    credentials,
    subscription_id
)

Registration

You also might need to add the “Monitoring Contributor Service Role” role to your credentials. See here to do it using the Python CLI: https://docs.microsoft.com/cli/azure/role/assignment

Get the Activity Log

This sample gets the logs from the ActivityLog of today for resource group ResourceGroupName, filtering the attributes to get only the “eventName” and “operationName”.

A complete list of available keywords for filters and available attributes is available here: https://msdn.microsoft.com/library/azure/dn931934.aspx

import datetime

today = datetime.datetime.now().date()
filter = " and ".join([
    "eventTimestamp ge {}".format(today),
    "resourceGroupName eq 'ResourceGroupName'"
])
select = ",".join([
    "eventName",
    "operationName"
])

activity_logs = client.activity_logs.list(
    filter=filter,
    select=select
)
for log in activity_logs:
    # assert isinstance(log, azure.monitor.models.EventData)
    print(" ".join([
        log.event_name.localized_value,
        log.operation_name.localized_value
    ]))

Metrics

This sample get the metrics of a resource on Azure (VMs, etc.).

A complete list of available keywords for filters is available here: https://msdn.microsoft.com/en-us/library/azure/mt743622.aspx

Supported metrics per resource type is available here: https://docs.microsoft.com/azure/monitoring-and-diagnostics/monitoring-supported-metrics

import datetime

# Get the ARM id of your resource. You might chose to do a "get"
# using the according management or to build the URL directly
# Example for a ARM VM
resource_id = (
    "subscriptions/{}/"
    "resourceGroups/{}/"
    "providers/Microsoft.Compute/virtualMachines/{}"
).format(subscription_id, resource_group_name, vm_name)

# You can get the available metrics of this specific resource
for metric in client.metric_definitions.list(resource_id):
    # azure.monitor.models.MetricDefinition
    print("{}: id={}, unit={}".format(
        metric.name.localized_value,
        metric.name.value,
        metric.unit
    ))

# Example of result for a VM:
# Percentage CPU: id=Percentage CPU, unit=Unit.percent
# Network In: id=Network In, unit=Unit.bytes
# Network Out: id=Network Out, unit=Unit.bytes
# Disk Read Bytes: id=Disk Read Bytes, unit=Unit.bytes
# Disk Write Bytes: id=Disk Write Bytes, unit=Unit.bytes
# Disk Read Operations/Sec: id=Disk Read Operations/Sec, unit=Unit.count_per_second
# Disk Write Operations/Sec: id=Disk Write Operations/Sec, unit=Unit.count_per_second

# Get CPU total of yesterday for this VM, by hour

today = datetime.datetime.now().date()
yesterday = today - datetime.timedelta(days=1)

filter = " and ".join([
    "name.value eq 'Percentage CPU'",
    "aggregationType eq 'Total'",
    "startTime eq {}".format(yesterday),
    "endTime eq {}".format(today),
    "timeGrain eq duration'PT1H'"
])

metrics_data = client.metrics.list(
    resource_id,
    filter=filter
)

for item in metrics_data:
    # azure.monitor.models.Metric
    print("{} ({})".format(item.name.localized_value, item.unit.name))
    for data in item.data:
        # azure.monitor.models.MetricData
        print("{}: {}".format(data.time_stamp, data.total))

# Example of result:
# Percentage CPU (percent)
# 2016-11-16 00:00:00+00:00: 72.0
# 2016-11-16 01:00:00+00:00: 90.59
# 2016-11-16 02:00:00+00:00: 60.58
# 2016-11-16 03:00:00+00:00: 65.78
# 2016-11-16 04:00:00+00:00: 43.96
# 2016-11-16 05:00:00+00:00: 43.96
# 2016-11-16 06:00:00+00:00: 114.9
# 2016-11-16 07:00:00+00:00: 45.4
# 2016-11-16 08:00:00+00:00: 57.59
# 2016-11-16 09:00:00+00:00: 67.85
# 2016-11-16 10:00:00+00:00: 76.36
# 2016-11-16 11:00:00+00:00: 87.41
# 2016-11-16 12:00:00+00:00: 67.53
# 2016-11-16 13:00:00+00:00: 64.78
# 2016-11-16 14:00:00+00:00: 66.55
# 2016-11-16 15:00:00+00:00: 69.82
# 2016-11-16 16:00:00+00:00: 96.02
# 2016-11-16 17:00:00+00:00: 272.52
# 2016-11-16 18:00:00+00:00: 96.41
# 2016-11-16 19:00:00+00:00: 83.92
# 2016-11-16 20:00:00+00:00: 95.57
# 2016-11-16 21:00:00+00:00: 146.73
# 2016-11-16 22:00:00+00:00: 73.86
# 2016-11-16 23:00:00+00:00: 84.7

Alert rules

This section shows how you can use the Python SDK to configure Azure metrics alerts. This enables you to automatically set up alerts on your resources when they are created to ensure that all resources are monitored correctly.

You need the azure-mgmt-monitor package for this feature.

You need to define three objects:

  • A data source
  • A condition on this data source
  • At least one action

There two possible data source:

For instance, to create a data source on a VM to alert on CPU usage:

from azure.mgmt.monitor.models import RuleMetricDataSource

resource_id = (
    "subscriptions/{}/"
    "resourceGroups/MonitorTestsDoNotDelete/"
    "providers/Microsoft.Compute/virtualMachines/MonitorTest"
).format(self.settings.SUBSCRIPTION_ID)

# I need a subclass of "RuleDataSource"
data_source = RuleMetricDataSource(
    resource_uri = resource_id,
    metric_name = 'Percentage CPU'
)

Supported metrics per resource type is available here: https://docs.microsoft.com/azure/monitoring-and-diagnostics/monitoring-supported-metrics

There is three possible conditions:

For instance, to create a threshold condition that triggers when the average CPU usage of a VM for the last 5 minutes is above 90% (using the preceding data source):

from azure.mgmt.monitor.models import ThresholdRuleCondition

# I need a subclasses of "RuleCondition"
rule_condition = ThresholdRuleCondition(
    data_source = data_source,
    operator = 'GreaterThanOrEqual',
    threshold = 90,
    window_size = 'PT5M',
    time_aggregation = 'Average'
)

There is two possible actions:

For instance, to create an email action:

from azure.mgmt.monitor.models import RuleEmailAction

# I need a subclass of "RuleAction"
rule_action = RuleEmailAction(
    send_to_service_owners = True,
    custom_emails = [
        'monitoringemail@microsoft.com'
    ]
)

Once you defined these thress objects that fit your needs, create the alert is a simple call:

rule_name = 'MyPyTestAlertRule'
my_alert = monitor_mgmt_client.alert_rules.create_or_update(
    group_name,
    rule_name,
    {
        'location': 'westus',
        'alert_rule_resource_name': rule_name,
        'description': 'Testing Alert rule creation',
        'is_enabled': True,
        'condition': rule_condition,
        'actions': [
            rule_action
        ]
    }
)