10ec68f493
Forty-one commits since v0.11.0. See CHANGELOG.md for the full list and the upgrade notes.
96 lines
3.3 KiB
TypeScript
96 lines
3.3 KiB
TypeScript
/**
|
|
* Health check endpoint — reports app and Immich connectivity status.
|
|
* Used by Docker HEALTHCHECK, monitoring tools, and load balancers.
|
|
*
|
|
* GET /api/health → { status, immich, uptime }
|
|
*/
|
|
|
|
import { NextRequest, NextResponse } from 'next/server';
|
|
import { immich } from '@/lib/immich';
|
|
import { checkRateLimit, getClientIp, retryAfterSeconds } from '@/lib/rate-limit';
|
|
import { env } from '@/lib/env';
|
|
import { getConfigOrNull } from '@/lib/config';
|
|
|
|
const startTime = Date.now();
|
|
|
|
// Cache the health check result for 10 seconds to prevent
|
|
// DoS attacks against the upstream Immich server while still
|
|
// allowing frequent polling from load balancers/orchestrators.
|
|
let cachedImmichOkPromise: Promise<boolean> | null = null;
|
|
let lastCheckTime = 0;
|
|
const CACHE_DURATION_MS = 10_000;
|
|
|
|
export async function GET(request: NextRequest) {
|
|
const ip = getClientIp(request);
|
|
// env, not getConfig(): a health check that cannot answer while the content
|
|
// config is unparsable is worse than useless — the Dockerfile probes this
|
|
// route, so a container with a broken gallery.yaml would restart in a loop at
|
|
// exactly the moment the app is still repairable through /admin (#519).
|
|
const rateLimitRpm = env.RATE_LIMIT_RPM;
|
|
|
|
const { success, remaining, resetAt } = checkRateLimit(`health:${ip}`, rateLimitRpm);
|
|
if (!success) {
|
|
return NextResponse.json(
|
|
{ error: 'Too many requests' },
|
|
{
|
|
status: 429,
|
|
headers: {
|
|
'Retry-After': String(retryAfterSeconds(resetAt)),
|
|
'X-RateLimit-Limit': String(rateLimitRpm),
|
|
'X-RateLimit-Remaining': String(remaining),
|
|
},
|
|
},
|
|
);
|
|
}
|
|
|
|
/*
|
|
* A config that cannot be parsed is not a reason to kill the container.
|
|
* Restarting cannot fix a YAML typo, the app is still serving the setup
|
|
* screen, and /admin — where the fix happens — is still reachable. Reporting
|
|
* it as unhealthy made the Dockerfile's HEALTHCHECK restart the container in
|
|
* a loop at exactly the moment it was still repairable (#519).
|
|
*
|
|
* Immich itself is not consulted here: the ping reads the same config and
|
|
* would answer "unreachable" for a reason that has nothing to do with Immich.
|
|
*/
|
|
if (!getConfigOrNull()) {
|
|
return NextResponse.json(
|
|
{
|
|
status: 'setup',
|
|
config: 'invalid',
|
|
immich: 'unknown',
|
|
uptime: Math.floor((Date.now() - startTime) / 1000),
|
|
},
|
|
{ status: 200, headers: { 'Cache-Control': 'no-store' } },
|
|
);
|
|
}
|
|
|
|
const now = Date.now();
|
|
if (!cachedImmichOkPromise || now - lastCheckTime > CACHE_DURATION_MS) {
|
|
cachedImmichOkPromise = immich.ping();
|
|
lastCheckTime = now;
|
|
}
|
|
|
|
// Await the potentially shared promise to prevent a thundering herd
|
|
// of concurrent requests from hitting the upstream server on a cache miss.
|
|
let immichOk = false;
|
|
try {
|
|
immichOk = await cachedImmichOkPromise;
|
|
} catch {
|
|
cachedImmichOkPromise = null; // reset on error so next request retries
|
|
immichOk = false;
|
|
}
|
|
|
|
const body = {
|
|
status: immichOk ? 'ok' : 'degraded',
|
|
config: 'ok',
|
|
immich: immichOk ? 'connected' : 'unreachable',
|
|
uptime: Math.floor((Date.now() - startTime) / 1000),
|
|
};
|
|
|
|
return NextResponse.json(body, {
|
|
status: immichOk ? 200 : 503,
|
|
headers: { 'Cache-Control': 'no-store' },
|
|
});
|
|
}
|