Skip to content
forked from google/ts-bridge

Time Series Bridge: a tool to import metrics from one monitoring system to another

License

Notifications You must be signed in to change notification settings

knyar/ts-bridge

 
 

Repository files navigation

Time Series Bridge is a tool that can be used to import metrics from one monitoring system into another. It regularly runs a specific query against a source monitoring system (currently only Datadog) and writes new time series results into the destination system (currently only Stackdriver).

ts-bridge is an App Engine Standard app written in Go.

Table of Contents

  1. Setup Guide
  2. metrics.yaml Configuration
  3. App Configuration
  4. Status Page
  5. Internal Monitoring
  6. Troubleshooting
  7. Development
  8. Support

Setup Guide

In brief, to set up the ts-bridge app:

  1. Create a GCP project that will host the app
  2. Configure metrics for import
  3. Deploy the app and let it auto-import your metrics every minute

The following sections will guide you through this process.

Create and Set Up a Google Cloud Project

We recommend making the project that hosts ts-bridge separate from the rest of your infrastructure so infrastructure failures will not affect monitoring and monitoring failues will not affect infrastructure.

  1. Log in to GCP and create a new Google Cloud project
  2. Ensure the new project is linked to a billing account (Note that the Stackdriver free tier can accommodate up to about 220 metrics. If you have already consumed your free quota with other usage, the incremental cost per metric is between US$2.32 and US$0.55 per month, depending on which pricing tier you are already in.)
  3. Enable stackdriver monitoring for the new project. When prompted:
    • Create a new stackdriver account for the project
    • Monitor only the new project (it should be selected by default)
    • Skip AWS setup and agent installation
    • Choose whether to receive email status reports for the project

Set Up A Dev Environment

We recommend using Cloud Shell to prepare ts-bridge for deployment to ensure a consistent and stable working environment. If you need a dev environment that you can share among multiple users, consider using a git repository and open-in-cloud-shell links.

  1. If you are not using Cloud Shell:
    • Install go
    • Download and install the Cloud SDK for Go
      • Initialize with the following commands to set the linked project and auth cookie:
      • gcloud init
      • gcloud auth application-default login
  2. Clone the ts-bridge source
    • go get github.com/google/ts-bridge/...
    • The ts-bridge source code should appear in ~/gopath/src/github.com/google/ts-bridge/

End To End Test (Dev Server)

  1. Ensure that you either have Owner permissions for the whole Cloud project, or at minimum the Monitoring Editor role

  2. Create a ts-bridge config with no metrics

    • cd ~/gopath/src/github.com/google/ts-bridge/app; cp metrics.yaml.example metrics.yaml
    • Edit the yaml file, remove the datadog_metrics sample content, and copy in the name of the project you just created into the stackdriver_destinations section.
    • Your metrics.yaml file should look like this:
    datadog_metrics:
    stackdriver_destinations:
      - name: stackdriver
        project_id: "your_project_name"
    
  3. Turn on the status page (uncomment #ENABLE_STATUS_PAGE: "yes" in app.yaml)

  4. Update SD_PROJECT_FOR_INTERNAL_METRICS in your app.yaml to match the name of your GCP project.

  5. Launch a dev server

    • dev_appserver.py app.yaml --port 18080
  6. Test via localhost/sync

    • curl http://localhost:18080/sync
  7. Verify that no error messages are shown. Troubleshooting guide:

    Error message Remedy
    ERROR: StatsCollector: rpc error: code = PermissionDenied desc = The caller does not have permission Ensure the authenticating user has at least the "Monitoring Editor" role
  8. Configure metrics by getting your API and application keys from Datadog and copying over your Datadog queries into metrics.yaml

    • Your metrics.yaml file should now look something like this:
    datadog_metrics:
      - name: your_first_metric_name
        query: "your metric query (copied from your Datadog dashboard)"
        api_key: aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
        application_key: bbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbb
        destination: stackdriver
      - name: your_second_metric_name
        query: "your metric query (copied from your Datadog dashboard)"
        api_key: aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
        application_key: bbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbb
        destination: stackdriver
    stackdriver_destinations:
      - name: stackdriver
        project_id: "your_project_name"
    
    • see below for configuration details
  9. Test metric ingestion via localhost/sync

    • curl http://localhost:18080/sync
  10. Verify that metrics are visible on status page

    • In Cloud Shell, click the ‘web preview’ button and change the port to 18080
    • If running on a local workstation, browse to http://localhost:18080/
  11. Verify that metrics are visible in the Stackdriver UI

  12. Kill the local dev server

  13. Revert SD_PROJECT_FOR_INTERNAL_METRICS to "" in app.yaml

Deploy In Production

  1. Ensure that you either have Owner permissions for the whole Cloud project, or at minimum the App Engine Admin and Cloud Scheduler Admin roles
  2. Disable the status page (comment out ENABLE_STATUS_PAGE: "yes" in app.yaml)
    • See below if you'd like to keep the status page enabled in prod.
  3. Create the App Engine application
    • gcloud app create
    • Choose the App Engine region. If you are using ts-bridge to import metrics originating from a system running on GCP, you should run ts-bridge in a different Cloud region from the system itself to ensure independent failure domains.
  4. Deploy app
    • gcloud app deploy --project <your_project_name> --version live
  5. Verify in the Stackdriver metrics explorer that metrics are being imported once a minute

metrics.yaml Configuration

Metric sources and targets are configured in the app/metrics.yaml file.

Metric Sources

Datadog

To import a metric from Datadog, ts-bridge regularly runs a configured query against the Datadog Query API.

Metrics imported from Datadog are defined in the datadog_metrics section of app/metrics.yaml. The following parameters need to be specified for each metric:

  • name: base name of the metric. While exporting to Stackdriver, this name will be prefixed with custom.googleapis.com/datadog/.
  • query: Datadog query expression. This needs to return a single time series (tags/labels are not supported yet).
  • api_key: Datadog API key.
  • application_key: Datadog Application key.
  • destination: name of the Stackdriver destination that query result will be written to. Destinations need to be explicitly listed in the stackdriver_destinations section of the configuration file.
  • cumulative: a boolean flag describing whether query result should be imported as a cumulative metric (a monotonically increasing counter). See Cumulative metrics section below for more details.

All parameters are required, except for cumulative (that defaults to false).

Please keep in mind the following details about Datadog API:

  • There is an API rate limit of 300 queries per hour that applies to the whole organization. Even if ts-bridge is the only user of the Query API, it still means you can only import 5 metrics if you are querying every minute (which is the default). The limit can be raised.
  • If you are using a rollup function as part of your query, Datadog will return a single point per each rollup interval. If rollup interval is longer than the importing period of ts-bridge, some import operations will fetch 0 new points. For example, if your query is producing a 10-minute ratio ( xxx.rollup(sum, 600) / yyy.rollup(sum, 600)) and you are using the default importing period (1 minute), ts-bridge will still issue the query every minute, however Datadog will only return a single point once every 10 minutes.
  • If you are not using the rollup function, Datadog will return points at maximum possible resolution (unless the query covers a very long time interval). Please keep in mind that Datadog might return more than 1 point per minute, and all points will be written to Stackdriver, even though Stackdriver does not allow querying with alignment period shorter than 1 minute.

Metric Destinations

Stackdriver

Imported metrics can be written to multiple destination Stackdriver projects, even though in practice we expect a single instance of Time Series Bridge to write to a single project (usually matching the GCP project where the ts-bridge is running).

Stackdriver destinations are listed in the stackdriver_destinations section of the app/metrics.yaml file. The following parameters can be specified for each destination:

  • name: name of the Stackdriver destination. It's only used internally by ts-bridge to match imported metrics with destinations.
  • project_id: name of the Stackdriver project that metrics will be written to. This parameter is optional; if not specified, the same project where ts-bridge is running will be used.

If you are using ts-bridge to write metrics to a different Stackdriver project from the one it's running in, you will need to grant roles/monitoring.editor IAM permission to the service account used by the ts-bridge App Engine app to allow it to read and write Stackdriver metrics.

App Configuration

Importing period

Time Series Bridge attempts to import all configured metrics regularly. This is driven by the App Engine Cron Service which is configured in app/cron.yaml. By default metrics are imported every minute.

Global settings

Some other settings can be set globally as App Engine environment variables via the env_variables section of app/app.yaml.

  • CONFIG_FILE: name of the metric configuration file (metrics.yaml).
  • SD_LOOKBACK_INTERVAL: time interval used while searching for recent data in Stackdriver. This is also the default backfill interval for when no recent points are found. This interval should be kept reasonably short to avoid fetching too much data from Stackdriver on each update.
    • You might be tempted to increase this significantly to backfill historic values. Please keep in mind that Stackdriver does not allow writing points that are more than 24 hours old. Also, Datadog downsamples values to keep the number of points in each response below ~300. This means that a single request can only cover a time period of 5 hours if you are aiming to get a point per minute.
  • UPDATE_TIMEOUT: the total time that updating all metrics is allowed to take. The incoming HTTP request from App Engine Cron will fail if it takes longer than this, and a subsequent update will be triggered again.
  • UPDATE_PARALLELISM: number of metric updates that are performed in parallel. Parallel updates are scheduled using goroutines and still happen in the context of a single incoming HTTP request, and setting this value too high might result in the App Engine instance running out of RAM.
  • DATADOG_MIN_POINT_AGE: minimum age of a data point returned by Datadog that makes it eligible for being written. Points that are very fresh (default is 1.5 minutes) are ignored, since Datadog might return incomplete data for them if some input data is delayed.
  • DATADOG_COUNTER_RESET_INTERVAL: while importing counters, ts-bridge needs to reset 'start time' regularly to keep the query time window small enough to avoid aggregation on Datadog side. This parameter defines how often a new start time is chosen. 30 minutes should be sufficient for metrics that have a point every 10 seconds. See Cumulative metrics section below for more details.
  • ENABLE_STATUS_PAGE: can be set to 'yes' to enable the status web page (disabled by default).

You can use --env_var flag to override these environment variables while running the app via dev_appserver.py.

Cumulative metrics

Stackdriver supports cumulative metrics, which are monotonically increasing counters. Such metrics allow calculating deltas and rates over different alignment periods.

While Datadog does not have first-class support for cumulative metrics, it is possible to use the cumsum query function to retreive a cumulative sum. Time Series Bridge can use result of such queries and import them as cumulative metrics, but such queries need to be explicitly annotated with a cumulative option in metrics.yaml being set to true.

For queries that are marked as cumulative, ts-bridge will regularly choose a 'start time' and then issue queries with that time passed in the from API parameter. As the result, Datadog will return a monotonically increasing time series with a sum of all measurements since 'start time'. To avoid aggregation of multiple points into one on Datadog side, 'start time' regularly gets moved forward, keeping the query time window short (see DATADOG_COUNTER_RESET_INTERVAL). Such resets are handled correctly by Stackdriver, since it requires explicit start time to be provided for cumulative metric points.

Often, for Datadog to provide a cumulative sum of all measurements, the .as_count() suffix needs to be appended to metric name. Otherwise measurements might be provided as per-second rates rather than exact counts.

For metrics that have measurements more often than every minute, you might also want to append the .rollup() function as described below.

For example, to import the counter metric called http_requests as a cumulative metric to Stackdriver, you might configure the following query in ts-bridge (and set cumulative to true):

cumsum(sum:http_requests{*}.as_count().rollup(sum, 60))

To unpack this:

  • cumsum() makes Datadog return a cumulative sum of measurements;
  • sum: prefix ensures that sum is used as the aggregation method if there are multiple time series with the same metric name but different tags (for example, reported from different machines);
  • .as_count() suffix gathers actual measurements rather than per-second rates;
  • .rollup(sum, 60) aggregates values into 60-second intervals in case there are multiple measurements for this metric reported per minute.

Status Page

If the ENABLE_STATUS_PAGE environment variable is set to 'yes', the index page of the App Engine app shows a list of configured metrics along with import status for each metric. This might be useful for debugging, however it is disabled by default to avoid publicly exposing a list of configured metrics (App Engine HTTP endpoints are publicly available by default).

If you choose to leave the status page enabled, we recommend configuring Identity-Aware Proxy (IAP) for the Cloud project in which ts-bridge is running. You can use IAP to restrict access to ts-bridge to a specific Google group or a list of Google accounts.

Internal Monitoring

Time Series Bridge uses OpenCensus to report several metrics to Stackdriver:

  • metric_import_latencies: per-metric import latency (in ms). This metric has a metric_name field.
  • import_latencies: total time it took to import all metrics (in ms). If this becomes larger than UPDATE_TIMEOUT, some metrics might not be imported, and you might need to increase UPDATE_PARALLELISM or UPDATE_TIMEOUT.
  • oldest_metric_age: oldest time since the last written point across all metrics (in ms). This metric can be used to detect queries that no longer return any data.

All metrics are reported as Stackdriver custom metrics and have names prefixed by custom.googleapis.com/opencensus/ts_bridge/

examples/ directory in this repository contains a suggested Stackdriver Alerting Policy you can use to receive alerts when metric importing breaks.

Troubleshooting

This section describes common issues you might experience with ts-bridge.

Writing points to Stackdriver too frequently

If your query returns more than 1 point per minute, you might be seeing the following error from Stackdriver:

One or more TimeSeries could not be written: One or more points were written more frequently than the maximum sampling period configured for the metric.

Stackdriver documentation recommends to not add points to the same time series faster than once per minute. If your Datadog query returns multiple points per minute, you can use the rollup function in your query to aggregate multiple points. For example, instead of a query like this:

sum:http_request_count{environment:prod}

You can use a query like this:

sum:http_request_count{environment:prod}.rollup(sum, 60)

In this example, rollup() will make sure the query returns a single point per minute, which will be a sum of all points within that minute.

Development

  • Set up a dev environment as per the Setup Guide above.
  • Create a metrics.yaml file in app/
  • Run the app locally using dev_appserver: cd app/ && dev_appserver.py app.yaml --port 18080
  • The app should be available at http://localhost:18080/
    • Note, dev_appserver does not support App Engine cron, so you'll need to run curl http://localhost:18080/sync to import metrics
  • Run tests: go test ./...
    • If you've changed interfaces, run go generate ./... to update mocks
  • If you've changed dependencies, run dep ensure to update vendored libraries and Gopkg.lock

If you'd like to contribute a patch, please see contribution guidelines in CONTRIBUTING.md.

Support

This is not an officially supported Google product.

About

Time Series Bridge: a tool to import metrics from one monitoring system to another

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Go 96.2%
  • HTML 3.8%