feat(analytics): instance census, full capture, richer error context (#511)

Add a once-per-boot instance_started event (arch, os, deploy_mode,
gpu_present) so the fleet architecture mix is measurable. It reuses the
existing per-instance instance_id and is exempt from the volume sample
rate, since a census that fires once per boot must not be thinned.

Restore useful capture depth now that the sponsored plan removes the
quota pressure behind the earlier hardening:

- PostHog sample rate 0.1 to 1.0 (full analytics when enabled); the
  property allowlist still blocks file data.
- Sentry per-instance ceiling 20 to 500/hr, breadcrumb trail restored
  (sanitized: urls/paths redacted, data payloads dropped), full stack
  paths kept; local vars, request bodies, and PII still dropped. Both
  api and web.

Honor ANALYTICS_ENABLED=false as an opt-out alias: it was documented on
the Docker Hub README but never wired in 2.x, so anyone who set it was
still tracked.

All capture stays behind the analytics opt-out gate.
This commit is contained in:
SnapOtter
2026-07-13 14:23:16 +08:00
committed by GitHub
parent b6fdabaea8
commit e1b8c24e5d
22 changed files with 378 additions and 106 deletions
+42 -7
View File
@@ -5,7 +5,10 @@
*/
import { rebuildErrorValue } from "@snapotter/shared";
const CEILING_PER_HOUR = 20;
// Per-process runaway guard, not a quota lever: under the sponsored plan we want
// real errors, but a single instance stuck in an error loop must not spam. Sentry
// de-dupes by fingerprint server-side, so 500 distinct events/hour is ample.
const CEILING_PER_HOUR = 500;
const HOUR_MS = 3600_000;
const TAG_ALLOWLIST = new Set([
@@ -21,9 +24,15 @@ const TAG_ALLOWLIST = new Set([
"status_code",
]);
function basename(p: string): string {
const i = Math.max(p.lastIndexOf("/"), p.lastIndexOf("\\"));
return i >= 0 ? p.slice(i + 1) : p;
const URL_RE = /https?:\/\/[^\s"')]+/g;
const BLOB_RE = /blob:[^\s"')]+/g;
// Absolute paths under roots that can hold user files (uploads live in /data,
// /tmp) or dev machines (/Users, /home). Over-redaction of a benign path is fine.
const PATH_RE = /(?:\/(?:Users|home|root|data|tmp|var|app|opt|mnt|srv)|[A-Za-z]:\\)[^\s"')]*/g;
/** Redact urls, blob refs, and absolute paths from free text (breadcrumb messages). */
function scrubText(s: string): string {
return s.replace(BLOB_RE, "<blob>").replace(URL_RE, "<url>").replace(PATH_RE, "<path>");
}
// Sentry event/hint are typed loosely on purpose: this module must not import
@@ -38,6 +47,30 @@ function asObj(value: unknown): AnyEvent | null {
: null;
}
// Keep the breadcrumb trail (the sequence of operations before the error) but
// strip content that can carry user data: redact urls/paths from the message and
// drop the structured `data` payload (http urls, query params) entirely.
function scrubBreadcrumb(entry: unknown): AnyEvent | null {
const b = asObj(entry);
if (!b) return null;
const out: AnyEvent = {};
for (const k of ["type", "category", "level", "timestamp"]) {
if (b[k] !== undefined) out[k] = b[k];
}
if (typeof b.message === "string") out.message = scrubText(b.message);
return out;
}
/** Sanitize the breadcrumb list, tolerating both the array and {values} shapes. */
function scrubBreadcrumbs(value: unknown): unknown {
if (Array.isArray(value)) return value.map(scrubBreadcrumb).filter(Boolean);
const wrapped = asObj(value);
if (Array.isArray(wrapped?.values)) {
return { values: wrapped.values.map(scrubBreadcrumb).filter(Boolean) };
}
return undefined;
}
export function buildBeforeSend(isActive: () => boolean) {
let windowStart = 0;
let sentInWindow = 0;
@@ -52,13 +85,15 @@ export function buildBeforeSend(isActive: () => boolean) {
}
if (++sentInWindow > CEILING_PER_HOUR) return null;
// Dropped: these surfaces can carry user data or PII.
event.message = undefined;
event.logentry = undefined;
event.server_name = undefined;
event.request = undefined;
event.extra = undefined;
event.breadcrumbs = undefined;
event.user = undefined;
// Kept, sanitized: the operation trail leading up to the error.
event.breadcrumbs = scrubBreadcrumbs(event.breadcrumbs);
const ctx = asObj(event.contexts);
const keep: AnyEvent = {};
@@ -88,8 +123,8 @@ export function buildBeforeSend(isActive: () => boolean) {
for (const entry of frames) {
const frame = asObj(entry);
if (!frame) continue;
if (typeof frame.filename === "string") frame.filename = basename(frame.filename);
frame.abs_path = undefined;
// Keep filename + abs_path (open-source code paths, not user data);
// drop only vars, which can hold user file contents or secrets.
frame.vars = undefined;
}
}