Monitoring health state changes

If you have configured application-based health checking for your managed instance group (MIG), Compute Engine writes a log entry whenever a managed instance's health state changes—for example, when the instance goes from HEALTHY to UNHEALTHY state. These log entries help you to monitor and debug the health state of each managed instance as well as the overall health of the MIG.

This document describes how to view and read health state change log entries, and provides specific use cases to help you monitor your managed instances.

Pricing

Compute Engine uses Cloud Logging to generate log entries for managed instance health state changes. Cloud Logging provides you with free allottment per month after which log entries are priced by data volume. For more information, see the Cloud Logging pricing summary.

To avoid the logging costs, you can disable the health state change logs.

Viewing health state change logs

Provided the health state change logs remain enabled, then Compute Engine writes a log entry to platform logs whenever the health state of a managed instance changes. You can view these logs for a project, for a specific MIG, or for a specific managed instance.

Viewing logs for a project or a MIG

To view log entries for a project or for a specific MIG, use the Cloud Console, gcloud tool, or the Compute Engine API.

Console

  1. Go to the Logs Explorer in the Google Cloud Console.

    Go to Logs Explorer

  2. Set the following query parameters:

    • Resource - GCE Instance Group Manager
    • Log name - instance_group_manager_events
  3. Alternatively, you can copy the following query in the Query builder.

    resource.type="gce_instance_group_manager" AND
    logName="projects/PROJECT_ID/logs/compute.googleapis.com%2Finstance_group_manager_events"
    

    You can narrow down your search to a specific managed instance group using the following query.

    resource.type="gce_instance_group_manager" AND
    logName="projects/PROJECT_ID/logs/compute.googleapis.com%2Finstance_group_manager_events" AND
    resource.labels.instance_group_manager_name="MIG_NAME"
    
  4. Click Run query. The Query results will display the logs.

gcloud

Use the gcloud logging read command to view and read the log entries.

To view all health state change logs in your project, use the following command:

gcloud logging read 'resource.type="gce_instance_group_manager" AND
    logName="projects/PROJECT_ID/logs/compute.googleapis.com%2Finstance_group_manager_events"' \
    --limit 10

To view all health state change logs for a specific managed instance group, use the following command:

gcloud logging read 'resource.type="gce_instance_group_manager" AND
    logName="projects/PROJECT_ID/logs/compute.googleapis.com%2Finstance_group_manager_events" AND
    resource.labels.instance_group_manager_name="MIG_NAME"' \
    --limit 10

Replace the following:

  • PROJECT_ID: your project id.
  • MIG_NAME: the MIG for which you want to view the health state change logs.

API

To view the health state change logs, make a POST request to the entries.list method.

To view all the health state change logs in your project, use the following command:

curl -H "Content-Type: application/json" -H "Authorization: Bearer OAUTH2_TOKEN" -X POST -d \
'{"filter":
    "projects/PROJECT_ID/logs/compute.googleapis.com%2Finstance_group_manager_events",
    "orderBy": "timestamp desc",
    "pageSize": 10,
    "resourceNames": ["projects/PROJECT_ID"]
}' https://logging.googleapis.com/v2/entries:list?alt=json

To view the health state change logs for a specific managed instance group, use the following command:

curl -H "Content-Type: application/json" -H "Authorization: Bearer OAUTH2_TOKEN" -X POST -d \
'{"filter":
    "resource.type=gce_instance_group_manager AND
    logName=projects/PROJECT_ID/logs/compute.googleapis.com%2Finstance_group_manager_events AND
    resource.labels.instance_group_manager_name=MIG_NAME",
    "orderBy": "timestamp desc",
    "pageSize": 10,
    "resourceNames": ["projects/PROJECT_ID"]
}' https://logging.googleapis.com/v2/entries:list?alt=json

Replace the following:

  • OAUTH2_TOKEN: your application's access token. For local testing, you can use the gcloud auth application-default print-access-token command to generate a token.
  • PROJECT_ID: your project id.
  • MIG_NAME: the MIG for which you want to view the health state change logs.

For more information about each log entry, see Format of log entries.

Depending on whether you want to archive the logs, use the logs for analysis, stream the logs to other applications, or trigger a Cloud Function, you can export the logs to destinations such as Cloud Storage, BigQuery, or Pub/Sub. For more information about exporting logs, see Overview of logs exports.

Viewing health state change logs for a specific VM

To view log entries for a specific managed instance, use the Cloud Console, gcloud tool, or the Compute Engine API.

Console

  1. Go to the Logs Explorer in the Google Cloud Console.

    Go to Logs Explorer

  2. Copy the following query in the Query builder.

    resource.type="gce_instance_group_manager" AND
    logName="projects/PROJECT_ID/logs/compute.googleapis.com%2Finstance_group_manager_events" AND
    jsonPayload.instanceHealthStateChange.instance=~"/INSTANCE_NAME$"
    
  3. Click Run query.

gcloud

Use the gcloud logging read command to view and read the log entries.

To view the health state change logs of a managed instance, use the following command:

gcloud logging read 'resource.type="gce_instance_group_manager" AND
    logName="projects/PROJECT_ID/logs/compute.googleapis.com%2Finstance_group_manager_events" AND
    jsonPayload.instanceHealthStateChange.instance=~"/INSTANCE_NAME$"' \
    --limit 10

Replace the following:

  • PROJECT_ID: your project id.
  • INSTANCE_NAME: the managed instance for which you want to view the health state change logs.

API

To view the health state change logs of a managed instance, make a POST request to the entries.list method.

curl -H "Content-Type: application/json" -H "Authorization: Bearer OAUTH2_TOKEN" -X POST -d \
'{"filter":
    "resource.type=gce_instance_group_manager AND
        logName=projects/PROJECT_ID/logs/compute.googleapis.com%2Finstance_group_manager_events AND
        jsonPayload.instanceHealthStateChange.instance=~\"/INSTANCE_NAME$\"",
    "orderBy": "timestamp desc",
    "pageSize": 10,
    "resourceNames": ["projects/PROJECT_ID"]
}' https://logging.googleapis.com/v2/entries:list?alt=json

Replace the following:

  • OAUTH2_TOKEN: your application's access token. For local testing, you can use the gcloud auth application-default print-access-token command to generate a token.
  • PROJECT_ID: your project id.
  • INSTANCE_NAME: the managed instance for which you want to view the health state change logs.

Format of log entries

Instance health state change log entries contain information useful for monitoring and debugging the state of your managed instances.

The logs are written to platform logs with the log name instance_group_manager_events. The platform logs help you to debug and troubleshoot issues.

logName="projects/PROJECT_ID/logs/compute.googleapis.com%2Finstance_group_manager_events"

Health state change log entries contain the following types of information:

  • General information shown in most logs, such as severity, project ID, MIG name and ID, project number, timestamp, and so on.
  • Fields specific to the instance's health state.

Within each health state change log entry, the jsonPayload.instanceHealthStateChange field contains the following information:

Field Description
instance URL for the instance, based on string project ID and instance name.
instanceWithId URL for the instance, based on its numeric project ID and instance ID.
ipAddress IP address of the instance, as probed by the health check.
network URL of the network resource for this instance, based on string project ID and network name.
networkWithId URL of the network resource for this instance, based on its numeric project ID and network ID.
healthCheck URL for the health check that's configured for the managed instance group.
previousDetailedHealthState Previous health state of the instance. For the list of possible states, see health states.
detailedHealthState Current health state of the instance. For the list of possible states, see health states.
notificationTime Timestamp of when the health state change occurred.

Log fields of type boolean typically only appear if they have a value of true. If a boolean field has a value of false, that field is omitted from the log.

UTF-8 encoding is enforced for log fields. Characters that are not UTF-8 characters are replaced with question marks.

Log entry example

The following example shows a VM instance's health state change from HEALTHY to UNHEALTHY:

  {
    "logName": "projects/my-project/logs/compute.googleapis.com%2Finstance_group_manager_events",
    "resource": {
      "type": "gce_instance_group_manager",
      "labels": {
        "instance_group_manager_id": "3138236342290985981",
        "instance_group_manager_name": "my-mig",
        "project_id": "my-project",
        "location": "europe-west3"
      }
    },
    "timestamp": "2019-11-19T15:47:57.127Z",
    "severity": "INFO",
    "jsonPayload": {
      "@type": "type.googleapis.com/compute.InstanceGroupManagerEvent",
      "instanceHealthStateChange": {
        "instance": "projects/my-project/zones/europe-west3/instances/my-mig-a",
        "instanceWithId": "projects/123456/zones/europe-west3/instances/6498902454451155884",
        "ipAddress": "10.0.0.4",
        "network": "projects/my-project/global/networks/net-1",
        "networkWithId": "projects/123456/global/networks/456",
        "healthCheck": "projects/my-project/global/healthChecks/my-mig-health-check",
        "previousDetailedHealthState": "HEALTHY",
        "detailedHealthState": "UNHEALTHY",
        "notificationTime": "2019-11-19T15:47:56.444Z"
      }
    },
    "receiveTimestamp": "2019-11-19T15:47:57.296439184Z"
  }
  

Use cases

You can use the health state change logs in the following monitoring or debugging scenarios:

  • Find out how often a particular VM instance changed health states over time.
  • Assess how often a MIG experienced health state changes of its instances.
  • Identify problematic VM instances that frequently go UNHEALTHY.
  • Find out what caused an autohealing attempt.
  • Find out whether an autohealing attempt succeeded for a specific VM instance.
  • Fine-tune health check configuration for an application by determining an appropriate initial delay for autohealing.

Monitoring health state changes of a VM

You can monitor how often a VM instance's health state changes by creating a metric that tracks the health state changes of that particular VM.

To create the metric and monitor the changes, do the following:

  1. Go to the Logs Explorer in the Google Cloud Console.

    Go to Logs Explorer

  2. Enter the following query in the Query builder, using your project ID and instance name.

    resource.type="gce_instance_group_manager" AND
    logName="projects/PROJECT_ID/logs/compute.googleapis.com%2Finstance_group_manager_events" AND
    jsonPayload.instanceHealthStateChange.instance=~"/INSTANCE_NAME$"
    
  3. In the Query results section, click Actions and then click Create metric.

  4. On the Create logs metric page, do the following:

    1. Set the Metric Type as Counter.
    2. Enter a Log metric name, for example health-mig-xyzq.

      The Build filter section displays the log query from the Logs Explorer. You can also configure the metric's filter to account for only disruptive states, such as UNHEALTHY and TIMEOUT, by adding severity>=WARNING to the filter.

    3. Under Labels, click Add label.

    4. Enter a Label name, for examplehealth_state.

    5. Set the Label type as STRING.

    6. Set the Field name to jsonPayload.instanceHealthStateChange.detailedHealthState. This will allow you to distinguish between different health state changes.

    7. Click Done to add the label.

    8. Click Create metric.

  5. Go to the Logs-based Metrics page and find the newly created metric.

  6. Click the menu in the metric's row and select View in Metrics Explorer. The Metrics explorer opens and displays the graph that represents the health state changes of the VM instance which you specified in the query.

Monitoring health state changes of all VMs in a MIG

You can monitor the health state changes of all managed instances by creating a metric that tracks the health state changes of the managed instances.

To create the metric and monitor the changes, do the following:

  1. Go to the Logs Explorer in the Google Cloud Console.

    Go to Logs Explorer

  2. Enter the following query in the Query builder, using your project ID and managed instance group's name.

    resource.type="gce_instance_group_manager" AND
    logName="projects/PROJECT_ID/logs/compute.googleapis.com%2Finstance_group_manager_events" AND
    jsonPayload.instanceHealthStateChange:* AND
    resource.labels.instance_group_manager_name="MIG_NAME"
    
  3. In the Query results section, click Actions and then click Create metric.

  4. On the Create logs metric page, do the following:

    1. Set the Metric Type as Counter.
    2. Enter a Log metric name, for example health-mig-xyzq.
    3. Under Labels, click Add label.
    4. Enter a Label name, for examplehealth_state.
    5. Set the Label type as STRING.
    6. Set the Field name to jsonPayload.instanceHealthStateChange.detailedHealthState. This will allow you to distinguish between different health state changes.
    7. Click Done to add the label.
    8. Click Create metric.
  5. Go to the Logs-based Metrics page and find the newly created metric.

  6. Click the menu in the metric's row and select View in Metrics Explorer. The Metrics explorer opens and displays the graph that represents the health state changes of all the VM instances in the managed instance group which you specified in the query.

Identifying VMs that frequently go unhealthy

You can identify problematic VMs that frequently go UNHEALTHY by creating a metric that tracks the health state changes of all VM instances in your MIG and grouping the metric by instances.

To create the metric and group by instances, do the following:

  1. Go to the Logs Explorer in the Google Cloud Console.

    Go to Logs Explorer

  2. Enter the following query in the Query builder, using your project ID and managed instance group's name.

    resource.type="gce_instance_group_manager" AND
    logName="projects/PROJECT_ID/logs/compute.googleapis.com%2Finstance_group_manager_events" AND
    jsonPayload.instanceHealthStateChange:* AND
    resource.labels.instance_group_manager_name="MIG_NAME"
    
  3. In the Query results section, click Actions and then click Create metric.

  4. On the Create logs metric page, do the following:

    1. Set the Metric Type as Counter.
    2. Enter a Log metric name, for example health-mig-xyzq.
    3. Under Labels, click Add label.
    4. Enter a Label name, for examplehealth_state.
    5. Set the Label type as STRING.
    6. Set the Field name to jsonPayload.instanceHealthStateChange.detailedHealthState. This will allow you to distinguish between different health state changes.
    7. Click Done to add the label.
    8. Similarly, add a second label, for example instance, with the Field name set to jsonPayload.instanceHealthStateChange.instance.
    9. Click Create metric.
  5. Go to the Logs-based Metrics page and find the newly created metric.

  6. Click the menu in the metric's row and select View in Metrics Explorer. The Metrics explorer opens and displays the graph that represents the health state changes of all the VM instances in the managed instance group which you specified in the query.

  7. Set the Group By to instance to see the aggregate number of health state changes for each instance.

The instances with the most health state changes in aggregate are the ones that most frequently go unhealthy.

Checking what caused an autohealing attempt

You can find out what caused an autohealing attempt by filtering logs for repair operations for given VM instance.

To filter the repair operations, do the following:

  1. Go to the Logs Explorer in the Google Cloud Console.

    Go to Logs Explorer

  2. Enter the following query in the Query builder, using your project ID and the instance's name.

    resource.type="gce_instance" AND
    logName="projects/PROJECT_ID/logs/cloudaudit.googleapis.com%2Fsystem_event" AND
    protoPayload.methodName="compute.instances.repair.recreateInstance" AND
    protoPayload.resourceName=~"/INSTANCE_NAME$"
    
  3. Click Run query. The Query results will show all autohealing attempts on the VM with the reason of autohealing in protoPayload.status.message.

Checking if autohealing succeeded for a VM

You can find out whether an autohealing attempt succeeded for a VM instance by filtering logs for repair operations and health changes by VM instance name. If the instance's health state changed to HEALTHY after a repair operation, you will see a corresponding health state change log. Follow the steps:

  1. Go to the Logs Explorer in the Google Cloud Console.

    Go to Logs Explorer

  2. Enter the following query in the Query builder, using your project ID and the instance's name.

    (resource.type="gce_instance" AND
    logName="projects/PROJECT_ID/logs/cloudaudit.googleapis.com%2Fsystem_event" AND
    protoPayload.methodName="compute.instances.repair.recreateInstance" AND
    protoPayload.resourceName=~"/INSTANCE_NAME$")
    OR
    (resource.type="gce_instance_group_manager" AND
    logName="projects/PROJECT_ID/logs/compute.googleapis.com%2Finstance_group_manager_events" AND
    jsonPayload.instanceHealthStateChange.instance=~"/INSTANCE_NAME$")
    

    The first part of the query displays the instance repair operations, which indicate that recreation was triggered by MIG autohealing to make the instance healthy again. The second part of the query displays all health state changes of the VM instance.

    In the Query results, the health state change event with detailedHealthState set to HEALTHY shortly after a repair operation shows that the autohealing attempt was successful.

Determining initial delay value of a MIG

Determining an appropriate initial delay value for MIG autohealing is easier with VM instance health state logging. You can use logs to observe the time between when the instances.insert operation finished and when the first healthy signal was received for a set of instances in a group. This time interval reveals how long instances take to fully boot up. As some VMs might boot up slower than others, Google recommends adding some margin to the observed initialization time (from insert operation to healthy state) when specifying the initial delay in the autohealing policy.

To measure the time between the instance insert operation and the instance becoming healthy, run a query for insert operations and health change logs by VM instance name. Use timestamps from both operations to calculate the instance's initialization time. Follow the steps:

  1. Go to the Logs Explorer in the Google Cloud Console.

    Go to Logs Explorer

  2. Enter the following query in the Query builder, using your project ID and the instance's name.

    (resource.type="gce_instance" AND
    logName:"projects/PROJECT_ID/logs/cloudaudit.googleapis.com%2Factivity" AND
    protoPayload.request.@type:"type.googleapis.com/compute.instances.insert" AND
    operation.last="true" AND
    "INSTANCE_NAME") OR
    (resource.type="gce_instance_group_manager" AND
    logName="projects/PROJECT_ID/logs/compute.googleapis.com%2Finstance_group_manager_events" AND
    jsonPayload.instanceHealthStateChange.instance=~"/INSTANCE_NAME$")
    

    The first part of the query shows the completion of the VM insert operation. The second part shows all health state changes for the VM.

    In the Query results, the timestamp of the health state change event with detailedHealthState set to HEALTHY shortly after the insert operation reveals the time needed for this VM to boot up.

  3. Repeat the steps for a few more VMs to get a better approximate value of the initial delay parameter.