mirror of
https://github.com/hanzoai/insights.git
synced 2026-10-30 16:15:31 +00:00
zeekay
10e6fec4c0
The rule table mapped HOG_ -> FN_ ahead of any POSTHOG rule, so POSTHOG_FOO split into POST + HOG_ and came out POSTFN_FOO. Nothing sets that. The live pods set INSIGHTS_SKIP_MIGRATION_CHECKS, which bin/migrate-check reads on the container entrypoint, so the misspelling does not degrade a feature -- it runs migrate --check against a database whose history this tree does not share and the container never starts. Both replacements are measured against the shipped tree rather than chosen: every HOG_ name with a counterpart there is spelled INSIGHTS_ (HOG_QL -> INSIGHTS_QL, HOG_FUNCTION_TEMPLATES -> INSIGHTS_FUNCTION_TEMPLATES, 12 of 12 determinate cases), and FN_ appears in it nowhere. Upstream carries 95 HOG_ tokens and zero FN_ ones, so repairing FN_ back out is exactly the inverse of the bad rule and cannot collide with an upstream name. The rule is fixed for the next convergence, and the tree it already produced is repaired in the same commit -- the codemod is idempotent against its own output, so it cannot undo this itself.
120 lines
5.5 KiB
TypeScript
120 lines
5.5 KiB
TypeScript
import { OnboardingComponentsContext, createInstallation } from 'scenes/onboarding/shared/OnboardingDocsContentWrapper'
|
|
|
|
import { StepDefinition } from '../steps'
|
|
|
|
export const getOpenTelemetrySteps = (ctx: OnboardingComponentsContext): StepDefinition[] => {
|
|
const { CodeBlock, Markdown, dedent } = ctx
|
|
|
|
return [
|
|
{
|
|
title: 'Scrape a Prometheus endpoint',
|
|
badge: 'recommended',
|
|
content: (
|
|
<>
|
|
<Markdown>
|
|
{dedent`
|
|
If your services already expose a Prometheus \`/metrics\` endpoint, run the Insights
|
|
metrics agent. It scrapes those endpoints and forwards them to Insights, so your
|
|
application code does not change.
|
|
`}
|
|
</Markdown>
|
|
<CodeBlock
|
|
blocks={[
|
|
{
|
|
language: 'bash',
|
|
file: 'Docker',
|
|
code: dedent`
|
|
docker run -d --name insights-metrics-agent \\
|
|
-e INSIGHTS_API_KEY=<ph_project_token> \\
|
|
-e INSIGHTS_HOST=<ph_client_api_host> \\
|
|
-e SCRAPE_TARGETS=your-app:9090,your-worker:9091 \\
|
|
insights/metrics-agent:latest
|
|
`,
|
|
},
|
|
]}
|
|
/>
|
|
<Markdown>
|
|
{dedent`
|
|
\`SCRAPE_TARGETS\` is a comma-separated list of \`host:port\` targets to scrape.
|
|
|
|
On Kubernetes, the Helm chart in \`products/metrics/agent/chart\` runs the same image.
|
|
Set \`shards\` to spread targets across a fleet, and \`persistence.enabled=true\` to keep
|
|
the export queue across restarts.
|
|
|
|
Counters and histograms scraped with OpenMetrics exemplars become clickable trace
|
|
links. Your Prometheus client has to attach the exemplars, the agent only preserves them.
|
|
`}
|
|
</Markdown>
|
|
</>
|
|
),
|
|
},
|
|
{
|
|
title: 'Send metrics over OTLP',
|
|
badge: 'optional',
|
|
content: (
|
|
<>
|
|
<Markdown>
|
|
{dedent`
|
|
If you instrument your application directly, Insights accepts metrics from any
|
|
OpenTelemetry client over OTLP HTTP. There is no Insights-specific package to install,
|
|
use the standard OpenTelemetry libraries for your language.
|
|
`}
|
|
</Markdown>
|
|
<CodeBlock
|
|
blocks={[
|
|
{
|
|
language: 'bash',
|
|
file: 'Environment variables',
|
|
code: dedent`
|
|
OTEL_EXPORTER_OTLP_METRICS_ENDPOINT="<ph_client_api_host>/i/v1/metrics"
|
|
OTEL_EXPORTER_OTLP_METRICS_HEADERS="Authorization=Bearer <ph_project_token>"
|
|
OTEL_SERVICE_NAME="my-app"
|
|
`,
|
|
},
|
|
{
|
|
language: 'yaml',
|
|
file: 'OTel Collector',
|
|
code: dedent`
|
|
receivers:
|
|
otlp:
|
|
protocols:
|
|
http:
|
|
endpoint: 0.0.0.0:4318
|
|
|
|
processors:
|
|
batch:
|
|
|
|
exporters:
|
|
otlphttp/insights:
|
|
metrics_endpoint: "<ph_client_api_host>/i/v1/metrics"
|
|
headers:
|
|
Authorization: "Bearer <ph_project_token>"
|
|
|
|
service:
|
|
pipelines:
|
|
metrics:
|
|
receivers: [otlp]
|
|
processors: [batch]
|
|
exporters: [otlphttp/insights]
|
|
`,
|
|
},
|
|
]}
|
|
/>
|
|
<Markdown>
|
|
{dedent`
|
|
The endpoint accepts OTLP over HTTP as \`application/x-protobuf\` or
|
|
\`application/json\`. gRPC is not supported, use the HTTP transport instead.
|
|
|
|
Gauges, sums, histograms, exponential histograms, and summaries are all accepted.
|
|
Keep label cardinality low: values like user IDs or request IDs create a new series
|
|
each time and get rate limited.
|
|
`}
|
|
</Markdown>
|
|
</>
|
|
),
|
|
},
|
|
]
|
|
}
|
|
|
|
export const OpenTelemetryInstallation = createInstallation(getOpenTelemetrySteps)
|