Skip to main content

Google Cloud Monitoring

Reading time: 0 minute(s) (0 words)

Google Cloud Monitoring (GCM) provides visibility into the performance, uptime, and overall health of cloud-powered applications. It collects metrics, events, and metadata from Google Cloud, hosted uptime probes, and application instrumentation.

Google Cloud Monitoring parameters and supported features in Nobl9
General support:
Release channel: Stable, Beta
Connection method: Agent, Direct
Replay and SLI Analyzer: Historical data limit 30 days
Event logs: Supported
Query checker: Not supported
Query parameters retrieval: Supported
Timestamp cache persistence: Supported
Query parameters:
Query interval: 1 min
Query delay: 2 min
Jitter: 15 sec
Timeout: 50 sec
Agent details and minimum required versions for supported features:
Plugin name: n9gcm
Query delay environment variable: GCM_QUERY_DELAY
Replay and SLI Analyzer: 0.80.0
Query parameters retrieval: 0.73.2
Timestamp cache persistence: 0.65.0
Additional notes:
Support for PromQL queries in Nobl9 agent versions 0.88.0 / 0.83.0-beta or later
Support for Google Cloud Monitoring metrics
Learn more Opens in a new tab

Creating SLOs with Google Cloud Monitoring​

Nobl9 Web​

Follow the instructions below to create your SLOs with Google Cloud Monitoring in the Nobl9 web application:

  1. Navigate to Service Level Objectives.

  2. Click .

  3. Select a Service.
    It will be the location for your SLO in Nobl9.

  4. Select your Google Cloud Monitoring data source.

  5. Enter Project ID: a unique identifier of your required Google Cloud project.

    • Can contain 6-30 lowercase letters, digits, or hyphens.
      For example, my-sample-project-191923
  6. Metric refers to the way of calculating and interpreting calculate and interpret data from your data source.
    • Threshold metric is defined by a single numerical value (the threshold) that separates satisfactory performance from unsatisfactory performance. It's represented by a single time series evaluated against the threshold.
    • Ratio metric expresses the performance as a fraction or proportion, typically by dividing the number of successful events by the total number of potential events (successes + failures). It's represented by two-time series for comparison for good events and total events.
      For ratio metrics, select the Data count method.

      SLI values for good and total
      When choosing the query for the ratio SLI (countMetrics), keep in mind that the values ​​resulting from that query for both good and total:
      • Must be positive.
      • While we recommend using integers, fractions are also acceptable.
        • If using fractions, we recommend them to be larger than 1e-4 = 0.0001.
      • Shouldn't be larger than 1e+20.

  7. Specify Query using PromQL.

    • Each query must return only one metric and one time series.
    • Since Nobl9 asks for data every 1 minute, we recommend setting the period for the align delta function to 1 minute, i.e. align delta(1m).
      As a result, Nobl9 receives the difference in a given minute and records it as an SLI.
    • Nobl9 processes a single dataset at a time and doesn't aggregate GCM metrics.
      Make sure your group_by aggregator points to the single dataset—exactly that one you want to observe.
      You can find the available groups on your Google Cloud Observability Monitoring dashboard > Metrics explorer.
    Click to open query samples
    Threshold metric
    {`sum(rate(serviceruntime_googleapis_com:api_request_latencies_sum{monitored_resource="consumed_api",service="bigquery.googleapis.com"}[1m]))/sum(rate(serviceruntime_googleapis_com:api_request_latencies_count{monitored_resource="consumed_api",service="bigquery.googleapis.com"}[1m]))`}
    Ratio metric, numerator
    {`sum(rate(serviceruntime_googleapis_com:api_request_count{monitored_resource=\"consumed_api\",response_code=\"200\",service=\"monitoring.googleapis.com\"}[1m]))`}
    Ratio metric, denominator
    {`sum(rate(serviceruntime_googleapis_com:api_request_count{monitored_resource=\"consumed_api\",service=\"monitoring.googleapis.com\"}[1m]))`}
    Monitoring Query Language

    MQL is no longer recommended by Google as a query language for Cloud Monitoring. Following this, MQL is deprecated in Nobl9 as well. PromQL is a recommended replacement.

    PromQL in GCM is supported by Nobl9 agent version 0.83.0-beta / 0.88.0 or higher.

    Refer to the PromQL Cheat Sheet for additional guidance.

  1. Define the Time window for your SLO:
  2. Configure the Error budget calculation method and Objectives:
    • Occurrences method counts good attempts against the count of total attempts.
    • Time Slices method measures how many good minutes were achieved (when a system operates within defined boundaries) during a time window.
    • You can define up to 12 objectives for an SLO.

      • Similar threshold values for objectives
      To use similar threshold values for different objectives in your SLO, we recommend differentiating them by setting varying decimal points for each objective.
      For example, if you want to use threshold value 1 for two objectives, set it to 1.0000001 for the first objective and to 1.0000002 for the second one.
  3. Add the Display name, Name, and other settings for your SLO:
    • Name identifies your SLO in Nobl9. After you save the SLO, its name becomes read-only.
      Use only lowercase letters, numbers, and dashes.
    • Select No data anomaly alert to receive notifications when your SLO stops reporting data for a specified period:
      • Choose up to five supported Alert methods.
      • Specify the delay period before Nobl9 sends an alert about the missing data.
        From 5 minutes to 31 days. Default: 15 minutes
    • Add alert policies, labels, and links, if required.
      Limits per SLO: 20 alert policies or links, 30 labels.
  4. Click CREATE SLO.

    5. SLO configuration use case
    Check the SLO configuration use case for a real-life SLO example.

YAML​

Monitoring Query Language

MQL is no longer recommended by Google as a query language for Cloud Monitoring. Following this, MQL is deprecated in Nobl9 as well. PromQL is a recommended replacement.

PromQL in GCM is supported by Nobl9 agent version 0.83.0-beta / 0.88.0 or higher.

Refer to the PromQL Cheat Sheet for additional guidance.

Monitoring Query Language is deprecated. We recommend using PromQL to write queries for Google Cloud Monitoring.

Sample GCM rawMetric SLO YAML definition with the MQL query (deprecated)
apiVersion: n9/v1alpha
kind: SLO
metadata:
  name: api-server-slo
  displayName: API Server SLO
  project: default
  labels:
    area:
      - latency
      - slow-check
    env:
      - prod
      - dev
    region:
      - us
      - eu
    team:
      - green
      - sales
  annotations:
    area: latency
    env: prod
    region: us
    team: sales
spec:
  description: Example Google Cloud Monitoring SLO
  indicator:
    metricSource:
      name: google-cloud-monitoring
      project: default
      kind: Agent
  budgetingMethod: Occurrences
  objectives:
    - displayName: Good response (200)
      value: 200
      name: ok
      target: 0.95
      rawMetric:
        query:
          gcm:
            query: |-
              fetch api-server
              | metric 'serviceruntime.googleapis.com/api/request_latencies'
              | filter (resource.service == 'monitoring.googleapis.com')
              | align delta(1m)
              | every 1m
              | group_by [resource.service],
                  [value_request_latencies_mean: mean(value.request_latencies)]
            projectId: my-project-id
      op: lte
      primary: true
  service: api-server
  timeWindows:
    - unit: Month
      count: 1
      isRolling: false
      calendar:
        startTime: '2022-12-01 00:00:00'
        timeZone: UTC
  alertPolicies:
    - fast-burn-5x-for-last-10m
  attachments:
    - url: https://docs.nobl9.com
      displayName: Nobl9 Documentation
  anomalyConfig:
    noData:
      alertMethods:
        - name: slack-notification
          project: default
      alertAfter: 1h
Click to open field reference
FieldTypeDescription
apiVersion
mandatory
stringAPI version. Use n9/v1alpha
kind
mandatory
stringThe resource type. Use SLO
Metadata
metadata.name
mandatory
stringName identifier for the SLO. Use only lowercase alphanumeric characters
metadata.displayNamestringUser-friendly SLO name
metadata.project
mandatory
stringThe name identifier of the project where you need to host your SLO
metadata.labelsobject (map: string[])Grouping labels for filtering or viewing
metadata.annotationsobject (map: string)Flat string annotations
Spec
spec.descriptionstringSLO description
spec.indicator.metricSource.name
mandatory
stringData source name
spec.indicator.metricSource.project
mandatory
stringProject containing the data source
spec.indicator.metricSource.kind
mandatory
stringData source connection method. Can be Agent or Direct
spec.budgetingMethod
mandatory
enumError budget calculation method. Can be Occurrences or Time slices
spec.objectives
mandatory
arrayYour SLO objective definition, up to 12 objectives per SLO.
spec.objectives[].displayNamestringUser-friendly objective name
spec.objectives[].value
mandatory
numberData point values that is considered "good" (e.g., 200.0).
In SLOs with two or more objectives, keep each objective's value unique.
In ratio (count) metrics, value is retained for legacy purposes.
spec.objectives[].name
mandatory
stringName identifier for this objective
spec.objectives[].op
mandatory
string (enum)Operator for objective. One of:
lte (less than or equal to)
lt (less than)
gte (greater than or equal to)
gt (greater than)
spec.objectives[].target
mandatory
floatThe percentage of the good minutes or occurrences that must meet the desired performance (e.g., is the target is 0.95, the good performance is expected to be observed in at least 95% of the time window)
spec.objectives[].rawMetric/.countMetric
mandatory
objectThe metric type indicator. Set:
rawMetric for a threshold metric
countMetric for a ratio metric.
A ratio metric requires the additional fields:
countMetric.incremental (boolean) the data count method
countMetric.good/.bad and countMetric.total a numerator and denominator queries
spec.objectives[].countMetric.incremental
mandatory
booleanThe data count method for a ratio (countMetric) metric type
spec.objectives[].primarybooleanThe indicator of a primary SLO objective
spec.service
mandatory
stringThe name identifier of a service to host this SLO. The service must exist in the project specified in metadata.project
spec.timeWindows
mandatory
arrayDefines SLO time window for error budget calculation. Set:
isRolling: true for the rolling time window type
isRolling: false for the calendar-aligned type
spec.timeWindows.unit
mandatory
integerThe time window units. One of:
Day | Hour | Minute for the rolling time window
Year | Quarter | Month | Week | Day for the calendar-aligned time window
spec.timeWindows.count
mandatory
integerThe number of units in a time window
spec.timeWindows.startTimestringMandatory for calendar-aligned time windows. Date and time in the format YYYY-MM-DDTHH:mm:ss
spec.timeWindows.timeZonestringMandatory for calendar-aligned time-windows. A valid IANA Time Zone Database name
spec.timeWindows.isRolling
mandatory
boolean
true for the rolling time window type
false for the calendar-aligned type
spec.alertPoliciesarrayThe name identifiers of alert policies to be linked to this SLO (must be from the same project as the SLO). Up to 20 alert policies per SLO.
spec.attachmentsarrayLinks to any additional attributes of this SLO
spec.anomalyConfigobjectSettings for a manual no data anomaly detection rule
spec.noData.alertMethodsarrayList of alert methods for no-data anomaly. Up to five alert methods per SLO. Every alert method must have the name and project fields
spec.noData.alertAfterstringWaiting time before sending a no-data notification. Must be 5m to 31d.
Default: 15m
Source-specific fields
gcm.query
mandatory
stringYour PromQL query
gcm.projectId
mandatory
stringA unique identifier of your required Google Cloud project. Must be a unique string of 6-30 lowercase alphanumeric characters, including hyphens (-)

Expected query output​

Nobl9 accepts single time series only. Therefore, at each point in the time series, the GCM query must return a single value.
When your query includes multiple tables, for example, using ident, make sure it returns a single value.
You can test your query result with the projects.timeSeries.query method

Sample expected query output returning a single value
{
  "timeSeriesDescriptor": {
    "pointDescriptors": [
      {
        "key": "good_total_ratio",
        "valueType": "DOUBLE",
        "metricKind": "GAUGE",
        "unit": "1"
      }
    ]
  },
  "timeSeriesData": [
    {
      "pointData": [
        {
          "values": [
            {
              "doubleValue": 0.9877300613496932
            }
          ],
          "timeInterval": {
            "startTime": "2024-06-06T08:00:03.532075Z",
            "endTime": "2024-06-06T08:00:03.532075Z"
          }
        }
      ]
    }
  ]
}

Querying the Google Cloud Monitoring server​

Google Cloud Monitoring API rate limits​

To verify the limits to API usage, go to the Quotas dashboard in the GCM UI. For an API, click the All Quotas button to see your quota.

Check out these related guides and references:
Add Google Cloud Monitoring as a data sourceAdding data sourcesprojects.timeSeries.queryGoogle Cloud docsCreating SLOs via TerraformTerraform
On this page