View fleet logs

This page describes how to enable and view logs for fleets. With fleet logging, multiple logs are aggregated and scoped together, enabling you to analyze the health of your applications in one consolidated view. This page is intended for:

  • Platform administrators who want to enable fleet logging and view logs in all namespaces.
  • Service operators who want to view logs in the specific namespaces to which they have access.

Overview

Fleet logs let you view logs at the entire fleet level, or for specific team scopes. Scopes are a team-management feature that let you define subsets of fleet logs and other resources on a per-team basis, with each scope associated with one or more fleet member clusters. For more information on scopes, see Manage teams for your fleet.

You can view two types of fleet logs:

  • Default logs: All Kubernetes logs (except Audit logs) that don't belong to any specific fleet scope with the following resource types:

    • k8s_container
    • k8s_pod
    • k8s_node
    • k8s_cluster
    • k8s_control_plane_components
  • Fleet scope logs: Container and Pod logs for applications owned by a team deployed in a specific fleet scope with multiple fleet-level namespaces.

Viewing fleet scope logs is optional. If you don't want to set up team management, you can still use fleet logging to view default logs.

Logs can be routed to different log buckets in the fleet host project with different views for access control. The default retention period of a log bucket is 30 days. You can configure this period if needed.

There are two modes supported for log routing where fleets contain clusters from multiple projects (cross-project registration):

  • MOVE: All logs are moved to the fleet host project. If a cluster in the fleet belongs to a different project, their logs are not retained in the original Google Cloud project.

  • COPY: All logs are sent to the fleet host project. If a cluster in the fleet belongs to a different project, their logs are also retained in the original Google Cloud project.

Before you begin

  1. If you have already manually created Cloud Logging buckets, sinks, and set exclusion filters, ensure that the names you have assigned these objects don't conflict with the fleet logging naming restrictions. If there is a naming conflict, contact Support before proceeding.

  2. Ensure that the clusters whose logs you want to view have been registered to your chosen fleet.

  3. If you don't have it installed already, install the Google Cloud CLI following the installation instructions. You need version 424.0.0 or higher to view your fleet logs.

  4. Ensure that your fleet host project has all the required APIs enabled, including the Anthos API:

    gcloud services enable --project=FLEET_HOST_PROJECT_ID  \
    gkehub.googleapis.com \
    container.googleapis.com \
    connectgateway.googleapis.com \
    cloudresourcemanager.googleapis.com \
    iam.googleapis.com \
    anthos.googleapis.com
    

    where:

Prepare scopes, namespaces and workloads

If you want to view fleet scope logs, you will need to create a fleet scope and a fleet namespace, in addition to preparing workloads for log collection.

Before you continue, set the default project for the Google Cloud CLI by running the following command:

gcloud config set project FLEET_HOST_PROJECT_ID

Create scopes and namespaces

If you want to view logs at the scope level, and haven't already set up scopes, follow the instructions in Manage teams for your fleet to create scopes, add clusters to scopes, and set up fleet namespaces.

Prepare workloads

To view log data from your applications, you will need to deploy your workloads in a cluster to the fleet namespace configured in the preceding step. This step is applicable whether you choose to view default logs, fleet scope logs, or both. Here is an example to configure your workload:

  apiVersion: v1
  kind: Pod
  metadata:
    name: fleet-example-pod
    namespace: NAMESPACE_NAME
  spec:
    containers:
    - name: count
      image: ubuntu:14.04
      args: [bash, -c,
           'for ((i = 0; ; i++)); do echo "$i: $(date)"; sleep 1; done']

After deploying the resource, you may see an error if the fleet namespace failed to create for some reason. In this case, run the following command to create the namespace again, and rerun the workload deployment command:

  kubectl create namespace NAMESPACE_NAME
  

Enable fleet logging

This section describes how to enable the fleet logging feature and grant team access to view logs.

gcloud

  • You can enable fleet logging using the Google Cloud CLI by specifying the configuration fields for the feature in a JSON or YAML file. Here is a example of a configuration for fleet logging in JSON format:

    {
      "loggingConfig": {
          "defaultConfig": {
              "mode": "COPY"
          },
          "fleetScopeLogsConfig": {
              "mode": "MOVE"
          }
      }
    }
    

To view all the fields you can configure for this feature, see the API reference.

When the defaultConfig or fleetScopeLogsConfig fields are enabled with the COPY or MOVE modes, as shown in the preceding example, a log sink is created with the prefix fleet-o11y-. This log sink is created under the Google Cloud project to route target logs from the cluster project to the fleet host project.

When fleetScopeLogsConfig is enabled, a log bucket with name fleet-o11y-scope-$SCOPE_NAME is also created in the global region under the fleet host project, if it doesn't exist already. Note that you can't change the bucket's region.

In this example, default logs will be sent to the fleet host project and retained in the original Google Cloud project, while fleet scope logs will be sent to the fleet host project, and not retained in the Google Cloud project.

  • Add your chosen configuration to a JSON file, and update the fleet:

    gcloud container fleet fleetobservability update \
            --logging-config=JSON_FILE
    

Replace JSON_FILE with the name of your filename.

Terraform

  • The fleet observability feature is enabled by default. If this is your first time using Terraform to manage the fleet observability feature, import the feature into Terraform by running the following command:
terraform import google_gke_hub_feature.feature projects/FLEET_HOST_PROJECT_ID/locations/global/features/fleetobservability

For example, you can add the following block to your Terraform configuration:

  resource "google_gke_hub_feature" "feature" {
    name = "fleetobservability"
    location = "global"
    spec {
      fleetobservability {
        logging_config {
          default_config {
            mode = "COPY"
          }
          fleet_scope_logs_config {
            mode = "MOVE"
          }
        }
      }
    }
  }

When the default_config or fleet_scope_logs_config fields are enabled with the COPY or MOVE modes, as shown in the preceding example, a log sink is created with the prefix fleet-o11y-. This log sink is created under the Google Cloud project to route target logs from the cluster project to the fleet host project.

When fleet_scope_logs_config is enabled, a log bucket with name fleet-o11y-scope-$SCOPE_NAME is also created under the fleet host project, if it doesn't exist already.

In this example, default logs will be sent to the fleet host project and retained in the original Google Cloud project, while fleet scope logs will be sent to the fleet host project, and not retained in the Google Cloud project.

Verify that the feature spec is updated:

   gcloud container fleet fleetobservability describe
  

The output shows the fleetobservability spec updated with the configuration, as in the following example:

createTime: '2022-09-30T16:05:02.222568564Z'
membershipStates:
  projects/123456/locations/us-central1/memberships/cluster-1:
    state:
      code: OK
      description: Fleet monitoring enabled.
      updateTime: '2023-04-03T20:22:51.436047872Z'
name:
projects/123456/locations/global/features/fleetobservability
resourceState:
  state: ACTIVE
spec:
  fleetobservability:
    loggingConfig:
      defaultConfig:
        mode: COPY
      fleetScopeLogsConfig:
        mode: MOVE
state:
  state: {}
updateTime: '2023-04-03T20:38:17.719596966Z'

Any changes made to the fleetobservability spec might take a few minutes to apply.

Set up cross-project logging permissions

This section is only required if you are registering a cluster to a fleet in a different project (also known as cross-project registration). In order to route logs from cluster projects to the fleet host project, you must grant the role roles/logging.bucketWriter to the logging service account from each cluster project.

  1. To obtain the service account credentials from sinks in cluster projects, run the following command:

    FLEET_HOST_PROJECT_ID=FLEET_HOST_PROJECT_ID
    FLEET_HOST_PROJECT_NUMBER=$(gcloud projects describe "${FLEET_HOST_PROJECT_ID}" --format "value(projectNumber)")
    gcloud logging sinks --project=GKE_PROJECT_ID describe fleet-o11y-${FLEET_HOST_PROJECT_NUMBER}-default
    

    If the command returns an error that the log sink cannot be found, try re-running the command after a minute or two. You can view the service account in the writerIdentity field of the sink description as shown in the following example:

    createTime: '2023-04-06T02:26:54.716195307Z'
    destination:
    logging.googleapis.com/projects/123456/locations/global/buckets/_Default
    filter: xxx
    name: fleet-o11y-default
    updateTime: '2023-04-06T19:03:51.598668462Z'
    writerIdentity:
    serviceAccount:service-123456@gcp-sa-logging.iam.gserviceaccount.com
    
  2. Grant the role of roles/logging.bucketWriter to the service account retrieved:

    gcloud projects add-iam-policy-binding FLEET_HOST_PROJECT_ID \
        --member "SERVICE_ACCOUNT" \
        --role "roles/logging.bucketWriter"
    

    where:

    • SERVICE_ACCOUNT is the name of the service account retrieved from the preceding step. For example:
    gcloud projects add-iam-policy-binding FLEET_HOST_PROJECT_ID \
        --member "serviceAccount:service-123456@gcp-sa-logging.iam.gserviceaccount.com" \
        --role "roles/logging.bucketWriter"
    

Grant team access to logs

The section describes how to grant access to users to view container logs and Pod logs.

  1. Get the IAM policy for the fleet project, and write it to a local file in JSON format:

    gcloud projects get-iam-policy FLEET_HOST_PROJECT_ID --format json > output.json
    
  2. Add an IAM condition that lets the user account view data from the log bucket you created. Here is an example to view container logs and Pod logs:

    {
      "bindings": [
        {
          "members": [
            "user:USER_ACCOUNT_EMAIL"
          ],
          "role": "roles/logging.viewAccessor",
          "condition": {
              "title": "Bucket reader condition example",
              "description": "Grants logging.viewAccessor role to user USER_ACCOUNT_EMAIL for the fleet-o11y-scope-SCOPE_NAME-k8s_container and fleet-o11y-scope-SCOPE_NAME-k8s_pod log view.",
              "expression":
                "resource.name == \"projects/FLEET_HOST_PROJECT_ID/locations/global/buckets/fleet-o11y-scope-SCOPE_NAME/views/fleet-o11y-scope-SCOPE_NAME-k8s_container\" || resource.name == \"projects/FLEET_HOST_PROJECT_ID/locations/global/buckets/fleet-o11y-scope-SCOPE_NAME/views/fleet-o11y-scope-SCOPE_NAME-k8s_pod\""
          }
        }
      ],
    }
    
  3. Update the IAM policy:

    gcloud projects set-iam-policy FLEET_HOST_PROJECT_ID output.json
    

For more options on granting access, see Grant access to a log view.

View fleet logs

Platform administrators have access to view all logs in all namespaces.

Default logs

To view all default logs in the _Default bucket in your fleet host project, fill in the variables in the following URL, copy and paste it to your browser:

https://console.cloud.google.com/logs/query;query=;storageScope=storage,projects%2FFLEET_HOST_PROJECT_ID%2Flocations%2Fglobal%2Fbuckets%2F_Default%2Fviews%2F_Default?jsmode=O&mods=pan_ng2&project=FLEET_HOST_PROJECT_ID

Fleet scope container logs and Pod logs

Service operators can view logs in the namespaces to which they have access. To view logs for all namespaces in a specific fleet scope, complete the following steps:

  1. With your fleet host project selected, go to the Teams section in the Google Cloud console.

    Go to Teams

  2. Click the team scope whose logs you want to view, and click the Logs tab.

  3. Select Container Logs or Pod logs to filter the logs view.

To view logs for a specific namespace in your scope:

  1. In the Teams page, with your team scope selected, click the Namespaces tab.
  2. Click the namespace whose logs you want to view, and click the Logs tab.
  3. Select either Container Logs or Pod logs to filter the logs view.

Alternatively, to view container logs, fill in the variables in the following URL, copy and paste it to your browser:

https://console.cloud.google.com/logs/query;query=;storageScope=storage,projects%2FFLEET_HOST_PROJECT_ID%2Flocations%2Fglobal%2Fbuckets%2Ffleet-o11y-scope-SCOPE_NAME%2Fviews%2Ffleet-o11y-scope-SCOPE_NAME-k8s_container?jsmode=O&mods=pan_ng2&project=FLEET_HOST_PROJECT_ID

To view Pod logs in a specific fleet scope, fill in the variables in the following URL, copy and paste it to your browser:

https://console.cloud.google.com/logs/query;query=;storageScope=storage,projects%2FFLEET_HOST_PROJECT_ID%2Flocations%2Fglobal%2Fbuckets%2Ffleet-o11y-scope-SCOPE_NAME%2Fviews%2Ffleet-o11y-scope-SCOPE_NAME-k8s_pod?jsmode=O&mods=pan_ng2&project=FLEET_HOST_PROJECT_ID

See Logs Explorer interface for more information on how to analyze log data.

Disable fleet logging

To disable the fleet logging feature, complete the following steps:

gcloud

  1. Save the following configuration to a file named disable_logging_config.json:

    {
      "loggingConfig": {}
    }
    
  2. Update the fleetobservability feature spec:

    gcloud container fleet fleetobservability update \
            --logging-config=disable_logging_config.json
    

Terraform

In your Terraform configuration, update all modes for log routing to MODE_UNSPECIFIED. Here is an example:

  resource "google_gke_hub_feature" "feature" {
    name = "fleetobservability"
    location = "global"
    spec {
      fleetobservability {
        logging_config {
          default_config {
            mode = "MODE_UNSPECIFIED"
          }
          fleet_scope_logs_config {
            mode = "MODE_UNSPECIFIED"
          }
        }
      }
    }
  }

Verify that the feature spec is updated:

   gcloud container fleet fleetobservability describe
  

The output shows the fleetobservability spec updated with your configuration:

  createTime: '2022-09-30T16:05:02.222568564Z'
  membershipStates:
    projects/123456/locations/global/memberships/cluster-1:
      state:
        code: OK
        description: Fleet monitoring enabled.
        updateTime: '2023-04-03T20:22:51.436047872Z'
  name:
  projects/123456/locations/global/features/fleetobservability
  resourceState:
    state: ACTIVE
  spec:
    fleetobservability:
      loggingConfig: {}
  state:
    state: {}
  updateTime: '2023-04-03T20:38:17.719596966Z'
  

Any changes made to the fleetobservability spec might take a few minutes to apply.

After disabling fleet logging, log sinks and exclusion filters will be removed from your projects. However, all log buckets created for the scope, and log views created under the log bucket will be preserved. To delete the log bucket in your fleet host project, see Delete a bucket.

Update retention period for log buckets

The default retention period of a log bucket is 30 days. To update this period, run the following command:

gcloud logging buckets update fleet-o11y-scope-SCOPE_NAME --location=global --retention-days=RETENTION_DAYS

where:

  • SCOPE_NAME is the name of the fleet scope.

  • RETENTION_DAYS is the number of days of the new retention period. For more options on configuring log buckets, see Manage buckets.

If you extend a bucket's retention period, then the retention rules apply going forward and not retroactively. Logs can't be recovered after the applicable retention period ends.

API reference

This section provides information on the possible fields you can add to your fleetobservability object.

fleetobservability

fleetobservability defines the fleet observability configuration.

Field Description Schema Optional
loggingConfig

Specified if the fleet logging feature is enabled for the entire fleet.

If unspecified, the fleet logging feature is disabled for the entire fleet.

loggingConfig True

loggingConfig

loggingConfig defines the configuration of fleet logging features in fleet observability.

Field Description Schema Optional
defaultConfig Sets the log routing behavior for default logs in the fleet. routingConfig True
fleetScopeLogsConfig Sets the log routing behavior for fleet scope logs. routingConfig True

routingConfig

routingConfig defines the configuration of the log routing mode in the fleet logging feature.

Field Description Schema Optional
mode

Specified to enable logs routing, and unspecified or MODE_UNSPECIFIED to disable logs routing.

If set to COPY, logs will be copied to the destination project.

If set to MOVE, logs will be moved to the destination project.

String; One of: MOVE, COPY and MODE_UNSPECIFIED True

Naming restrictions

When fleet observability is enabled, the fleet observability controller reserves the following names for the logs objects it creates. To avoid unwanted or unexpected behavior, you should avoid using these names when you create your own log buckets, sinks, and set exclusion filters.

Feature enabled Object created Name used by fleet observability
defaultConfig Sink fleet-o11y-FLEET_PROJECT_NUMBER-default
Exclusion filter. fleet-o11y-FLEET_PROJECT_NUMBER-default-exclusion. This name is reserved under the _Default sink of the cluster project.
fleetScopeLogsConfig Log bucket fleet-o11y-scope-SCOPE_NAME
  • Logs view for container logs in the bucket
  • fleet-o11y-scope-SCOPE_NAME-k8s_container
  • Logs view for Pod logs in the bucket
  • fleet-o11y-scope-SCOPE_NAME-k8s_pod
    Sink fleet-o11y-FLEET_PROJECT_NUMBER-scope-SCOPE_NAME
    Exclusion filter fleet-o11y-FLEET_PROJECT_NUMBER-scope-exclusion

    Troubleshooting

    This section describes how to resolve fleet logging related issues.

    Email notification about sink configuration error

    If you received an email with the title [ACTION REQUIRED] Cloud Logging sink configuration error in <Your GCP Project>, then the service account of your log sink does not have permission to write logs to the destination of the sink. To resolve this, follow the steps in Cross-project logging permissions.

    Unknown error message from Cloud Logging UI

    If you see the following error in the Cloud Logging UI, double-check that the project_id and scope variables entered in the URL are correct.

    Error: There is an unknown error while executing this operation.
    

    Membership not found error

    You might see the following error:

    ERROR: (gcloud.alpha.container.fleet.memberships.bindings.create) NOT_FOUND: Resource 'parent resource not found for projects/...' was not found
    

    Ensure that you have registered the cluster to a fleet.