Health check logging information

You can get logs of load balancing health checks when an endpoint changes in health status. You can use health check logs for the following:

  • Live debugging and troubleshooting your endpoint health status
  • Getting visibility into your endpoint's health status
  • Auditing and compliance purposes

Health checks log health transition information to Logging. You enable or disable logging on a per-health check basis.

To see health check logs in Logging, you must ensure that you don't have a logs exclusion that applies to health checks. For instructions about how to verify that GCE Instance Group and Network Endpoint Group logs are allowed, see Exclusion filters.

Enable and disable logging

This section describes how to enable logging on a new or an existing health check, and how to disable logging on an existing health check.

Enable logging on a new health check

Console

  1. In the Google Cloud console, go to the Health checks page.

    Go to Health checks

  2. Click Create health check.

  3. For Logs, select On.

  4. Continue setting up your health check.

gcloud

gcloud compute health-checks create PROTOCOL HEALTH_CHECK_NAME \
    --enable-logging

The flag --enable-logging enables logging for that health check.

Terraform

To create a health check for different protocols with logging, use the google_compute_health_check resource.

resource "google_compute_health_check" "health_check_tcp_with_logging" {
  provider = google-beta

  name = "health-check-tcp"

  timeout_sec        = 1
  check_interval_sec = 1

  tcp_health_check {
    port = "22"
  }

  log_config {
    enable = true
  }
}

For a regional load balancer, use the google_compute_region_health_check resource.

To learn how to apply or remove a Terraform configuration, see Basic Terraform commands.

Enable logging on an existing health check

Console

  1. In the Google Cloud console, go to the Health checks page.

    Go to Health checks

  2. Click the name of your health check.

  3. Click Edit.

  4. For Logs, select On.

  5. Click Save.

gcloud

gcloud compute health-checks update PROTOCOL HEALTH_CHECK_NAME \
    --enable-logging

The flag --enable-logging enables logging for that health check.

Disable logging on an existing health check

Console

  1. In the Google Cloud console, go to the Health checks page.

    Go to Health checks

  2. Click the name of your health check.

  3. Click Edit.

  4. For Logs, select Off.

  5. Click Save.

gcloud

gcloud compute health-checks update PROTOCOL HEALTH_CHECK_NAME \
    --no-enable-logging

The flag --no-enable-logging disables logging for that health check.

View logs

  1. To view logs, go to the Logs explorer.

    Health check logs are indexed by instance group or network endpoint group.

  2. To see all logs, in the Resource menu, select either GCE Instance Group or Network Endpoint Group, depending on the type of backend.

  3. Alternatively, paste the following into the Query field. Replace PROJECT_ID with your project's ID.

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

You can configure export of logs-based metrics for load balancer health checks.

Use filters to view logs

You can also get logs based on more specific searches. For example, the following filter shows all logs for a specified backend instance IP address:

  logName="projects/PROJECT_ID/logs/compute.googleapis.com%2Fhealthchecks"
jsonPayload.healthCheckProbeResult.ipAddress="IP_ADDRESS"

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.

What is logged

Health check log entries contain information useful for monitoring and debugging the state of your endpoints. Log entries contain the following types of information:

  • General information shown in most logs, such as severity, project ID, project number, and timestamp.
  • Fields specific to health checks, described in the following tables.

Health check states

An endpoint is considered either HEALTHY or UNHEALTHY. These are the basic states. Within each of these basic states, there are several more detailed states.

Hybrid NEGs and regional internet NEGs using distributed Envoy health checks do not support detailed health states.

The following table shows the mapping between basic and detailed health states.

Basic health state Detailed health state
HEALTHY HEALTHY
DRAINING
UNHEALTHY UNKNOWN
UNHEALTHY
TIMEOUT

State changes don't always alter the behavior of the load balancer. Consider the following case:

  1. The server is giving the wrong response, so the endpoint is considered UNHEALTHY.
  2. The server then stops responding, and the new state is TIMEOUT.
  3. The load balancer still considers the endpoint to be UNHEALTHY because the detailed TIMEOUT state maps to the basic UNHEALTHY state.

The following table provides a definition of each health state.

Detailed health check state Meaning Basic state
HEALTHY The endpoint is reachable and conforms to the requirements defined by the health check. HEALTHY
UNHEALTHY The endpoint is reachable but does not conform to the requirements defined by the health check. UNHEALTHY
DRAINING The endpoint is being drained. The existing connections to the endpoint are allowed to complete, but the new ones are being refused. Endpoint is considered HEALTHY. HEALTHY
TIMEOUT The endpoint is unreachable. Depending on the health check type, either a connection to the endpoint cannot be established or the server did not respond within the specified timeout. Endpoint is considered UNHEALTHY. UNHEALTHY
UNKNOWN The health check system is aware of the endpoint, but its health is not known. Endpoint is considered UNHEALTHY. UNHEALTHY

There are multiple health checkers probing each endpoint; Google Cloud de-duplicates the log entries before logging so that only unique logs are generated.

If a health checker restarts, you might occasionally see the logged health state change from UNKNOWN to one of the known states listed earlier even though the endpoint's health state has not actually changed. Google Cloud uses best-effort heuristics to suppress such log entries.

If you are using connection draining, then health check logs are not be generated with the endpoint health status DRAINING. This is because health check logs reflect the results observed by the health check probes, and connection draining does not impact the results observed by the health check probe. Connection draining works by just informing the load balancer that the new state is DRAINING, and it effectively overrides the true health state of the endpoint as observed by the health checker.

You can interact with the logs by using the Cloud Logging API. The API provides ways to interactively filter logs that have specific fields set and export matching logs to Cloud Logging, Cloud Storage, BigQuery, or Pub/Sub. For more information about the Cloud Logging API, see Cloud Logging API overview.

Health check log entry

The LogEntry jsonPayload is populated with a field healthCheckProbeResult that contains the following information.

Field Type Description
ipAddress string The primary internal IP address associated with the primary network interface of each backend VM. This is a human-readable string.
healthCheckProtocol enum(HealthCheckProtocol) The health check protocol used for health checking the endpoint. Examples: TCP, HTTP, HTTPS.
healthState enum(HealthState) Current health status of the endpoint: HEALTHY or UNHEALTHY.
previousHealthState enum(HealthState) The previous health status of the endpoint: HEALTHY or UNHEALTHY.
detailedHealthState enum(DetailedHealthState) Current detailed health status of the endpoint. For a list of possibilities, see Health check states.

Not supported for distributed Envoy health checks for hybrid NEGs and regional internet NEGs.

previousDetailedHealthState enum(DetailedHealthState) The previous detailed health status of the endpoint. For a list of possibilities, see Health check states.

Not supported for distributed Envoy health checks for hybrid NEGs and regional internet NEGs.

probeRequest string

For HTTP, HTTPS, and HTTP/2, this is the URL request path (requestPath field in the resource config).

For TCP/SSL, this is the configured optional string that is sent after the health check connection is established (request field in the resource config).

Not supported for distributed Envoy health checks for hybrid NEGs and regional internet NEGs.

probeCompletionTimestamp google.protobuf.Timestamp Probe completion timestamp.
connectLatency google.protobuf.Duration Time spent on setting up the connection for connection-oriented health check protocols TCP, SSL, HTTP, HTTPS, and HTTP/2.

Not supported for distributed Envoy health checks for hybrid NEGs and regional internet NEGs.

responseLatency google.protobuf.Duration Latency between request and response, as measured by the prober.

Not supported for distributed Envoy health checks for hybrid NEGs and regional internet NEGs.

probeResultText string Descriptive text related to the probe result. It might say something like "Connection timed out" or "HTTP response: Bad Gateway", or it could be empty.

Not supported for distributed Envoy health checks for hybrid NEGs and regional internet NEGs.

probeSourceIp string The IP address from which the health check probe was sent.

For distributed Envoy health checks, this corresponds to the proxy IP address from the proxy-only subnet.

probeSourceRegion string

The cloud vantage point region from which the health check originated.

This field is populated only for probes of Cloud DNS routing policies if the sourceRegions field is set in the corresponding health check.

targetIp string The IP address that is the target of the probe. This can be different than ipAddress. The destination IP address of the probe depends on the type of load balancer. For details, see Destination for probe packets in the Health checks overview.
targetPort int The port that was the target of the probe. This can be the default port of the probe or the port that you specified when you created the health check.

Example filters

This section provides examples of common log filters.

Find all health check results for a particular instance group

logName="projects/PROJECT_ID/logs/compute.googleapis.com%2Fhealthchecks"  AND
resource.type="gce_instance_group" AND
resource.labels.instance_group_name="INSTANCE_GROUP_NAME"

Find all health check results for a particular NEG

logName="projects/PROJECT_ID/logs/compute.googleapis.com%2Fhealthchecks"  AND
resource.type="gce_network_endpoint_group" AND
resource.labels.network_endpoint_group_id="ENDPOINT_GROUP_ID"

Find all health check transitions for backend instance IP address 10.128.15.201

logName="projects/PROJECT_ID/logs/compute.googleapis.com%2Fhealthchecks"
jsonPayload.healthCheckProbeResult.ipAddress="10.128.15.201"

Find all endpoints that were previously HEALTHY but now TIMEOUT

logName="projects/PROJECT_ID/logs/compute.googleapis.com%2Fhealthchecks"
jsonPayload.healthCheckProbeResult.previousDetailedHealthState="HEALTHY"
jsonPayload.healthCheckProbeResult.detailedHealthState="TIMEOUT"

Find health logs from a specific time range

logName="projects/PROJECT_ID/logs/compute.googleapis.com%2Fhealthchecks"
timestamp>"2019-02-14T02:20:00.0Z"
timestamp<"2019-02-14T03:30:00.0Z"

Limitations

  • Logs are generated for endpoint health transition only.
  • Legacy health checks are not supported.
  • Target pools are not supported.
  • Logs are not generated when the endpoint's health state is UNKNOWN.
  • In case of VM migrations, you might not see any log entries generated when the endpoint's health transitions to the UNHEALTHY state.
  • Logs are not generated when endpoints are deleted. For example, when you stop a VM.

What's next