[DOCS-13132] Clarify Google Cloud metric ingestion latency#36313
[DOCS-13132] Clarify Google Cloud metric ingestion latency#36313buraizu wants to merge 2 commits into
Conversation
Preview links (active after the
|
| {{% /collapse-content %}} | ||
|
|
||
| Metrics appear in Datadog approximately **15 minutes** after setup. | ||
| The first metrics appear in Datadog approximately **15 minutes** after setup. After the initial setup, expect a steady-state delay of approximately **7-8 minutes** for new metrics. See [Cloud Metric Delay][412] for details. |
There was a problem hiding this comment.
The p99 on time-to-first-point for GCP is now ~2 minutes for the vast majority of metrics, so I think we could update the wording here to reflect that.
There was a problem hiding this comment.
Applied — updated initial-setup latency from 15 minutes to 2 minutes in bdebf1c
|
|
||
| Google Cloud emits metrics with one-minute granularity, and Datadog pulls them through the Google Cloud Monitoring API. Expect a steady-state delay of approximately 7-8 minutes between when Google Cloud records a metric and when it appears in Datadog. See [Cloud Metric Delay][7] for the full breakdown across cloud providers. | ||
|
|
||
| When troubleshooting a metric that appears delayed but not missing, wait at least 10-15 minutes before assuming it is missing. For initial integration setup, the first metrics appear in Datadog approximately 15 minutes after the integration is configured. |
There was a problem hiding this comment.
The various metrics pages linked from this page have more info on a per-metric basis, but there some metrics that have slow sample rate or high delay from GCP's end. So I think we could link to those GCP docs as well, since we can't do anything about high metric delay if the delay is from GCP
There was a problem hiding this comment.
Applied — added a sentence linking to the Google Cloud metric catalog for per-metric timing info, noting that Datadog cannot reduce delays within Google Cloud in bdebf1c
2 participants
What does this PR do? What is the motivation?
Fixes DOCS-13132
Clarifies the Google Cloud metric ingestion latency in two places to deflect "why am I waiting" support tickets analyzed under DOCS-13132. The existing "Metrics appear in Datadog approximately 15 minutes after setup" wording conflated initial-setup latency (time before the first metric arrives after configuring the integration) with steady-state latency (ongoing delay between Google Cloud emitting a metric and it appearing in Datadog).
Changes:
content/en/getting_started/integrations/google_cloud.md: distinguishes the ~15-minute initial-setup latency from the ~7-8-minute steady-state delay, with a reference link to the existing Cloud Metric Delay guide for the full breakdown.content/en/integrations/guide/gcp-metric-discrepancy.md: adds a new## Latencysection before## Further reading, covering the same two numbers plus a "wait at least 10-15 minutes before assuming a metric is missing" troubleshooting guard.The 7-8 minute number is sourced from the existing
cloud-metric-delayguide, which has been the canonical Datadog-wide cloud latency reference. This PR surfaces that data point in the GCP-specific docs where customers actually search.A companion change in
integrations-internal-coreapplies the same disambiguation to the integration tile page (google_cloud_platform/README.md); that PR will be opened after this one merges.Follow-up work tracked in DOCS-14214: log latency, billing latency, and backfill behavior across pipelines (SME-blocked).
Merge instructions
Merge readiness:
Additional notes