ImageMagick sizes its pixel cache against the machine and will hold an entire decoded image in RAM. On a host with room that is right. On a small one it is how resizing a photograph kills whatever else is running — and "whatever else" includes the website this service exists to serve, which is not an acceptable way for an image to fail. So the decode runs under `-limit memory` / `-limit map` / `-limit disk`. Past the caps ImageMagick spills to a temporary file and carries on, which turns an out-of-memory kill of whatever the kernel picks into a conversion that is slower. Nobody minds a slow photograph; everybody minds a site that went down while one was being resized. Measured on fofos. A 12 MP HEIC peaks at 145 MB, so the 256 MB default does not bind and the output is byte-identical — it is a ceiling, not a tax. Forced to bind, on a 36 MP image with a 32 MiB cap, peak RSS is 12 MB and it still exits 0 with the file written. `-limit thread 1` for the same reason: one decode is one image, and each thread carries its own share of the cache the caps exist to bound. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
118 lines
5.4 KiB
TypeScript
118 lines
5.4 KiB
TypeScript
import { resolve } from "node:path";
|
||
|
||
/**
|
||
* Configuration, read once at boot.
|
||
*
|
||
* Every limit here exists because its absence is a denial of service — see docs/SECURITY.md,
|
||
* "Layer 5 — bounded work". They are configurable so a deployment can tighten them, not so a
|
||
* deployment can remove them: each has a floor enforced below.
|
||
*/
|
||
|
||
function num(key: string, fallback: number, min: number): number {
|
||
const raw = process.env[key];
|
||
if (raw === undefined) return fallback;
|
||
|
||
const value = Number(raw);
|
||
if (!Number.isFinite(value)) {
|
||
throw new Error(`${key} must be a number, got ${JSON.stringify(raw)}`);
|
||
}
|
||
if (value < min) {
|
||
throw new Error(`${key} must be at least ${min}, got ${value}`);
|
||
}
|
||
return value;
|
||
}
|
||
|
||
/**
|
||
* The port to listen on. **0 means "any free one", and that is the default.**
|
||
*
|
||
* A plugin sidecar does not choose a number. It binds whatever is free and REPORTS it to the platform,
|
||
* which is how every other sidecar here works and why none of them can collide. av-club asserted 3020
|
||
* because it began life as a standalone service, and on 2026-08-28 that assumption cost a deploy: a
|
||
* stale copy of the pre-plugin code was still holding 3020 on fofos, so the plugin's sidecar died on
|
||
* EADDRINUSE in a restart loop.
|
||
*
|
||
* `AV_CLUB_PORT` still works and still pins the port. That is for running this outside Officer — the
|
||
* standalone mode fofos/platform was written against — where something on the other side has to know
|
||
* where to find it. Inside Officer nothing should set it.
|
||
*/
|
||
export const PORT = num("AV_CLUB_PORT", 0, 0);
|
||
|
||
/**
|
||
* Where assets live. Outside the repo by default, and deliberately not under any directory a deploy
|
||
* would replace — losing every stored asset because the working tree was refreshed is the kind of
|
||
* failure that is only noticed later, by which point the originals are gone.
|
||
*/
|
||
export const ASSETS_DIR = resolve(process.env.AV_CLUB_ASSETS_DIR ?? `${process.env.HOME}/.local/share/av-club/assets`);
|
||
|
||
/** Multipart bodies are buffered whole. Without a cap, one request is an out-of-memory. */
|
||
/**
|
||
* The only part of the filesystem this service will read from or write to when a caller names a path.
|
||
*
|
||
* Defaults to the owner's home, which is where every consumer's files already are — the file browser
|
||
* runs there, and so does the commander watching an upload folder. Narrow it with the env var on a
|
||
* machine where av-club should only see one project.
|
||
*
|
||
* `/` is refused outright rather than treated as "no limit": a root that permits everything is a
|
||
* configuration mistake that would look like it was working.
|
||
*/
|
||
export const WORK_ROOT = (() => {
|
||
const root = resolve(process.env.AV_CLUB_WORK_ROOT ?? process.env.HOME ?? "");
|
||
if (!root || root === "/") {
|
||
throw new Error("AV_CLUB_WORK_ROOT (or HOME) must be a real directory, and may not be /");
|
||
}
|
||
return root;
|
||
})();
|
||
|
||
export const MAX_UPLOAD_BYTES = num("AV_CLUB_MAX_UPLOAD_BYTES", 512 * 1024 * 1024, 1024);
|
||
|
||
/**
|
||
* How much RAM ImageMagick may hold while decoding, before it spills to a temporary file.
|
||
*
|
||
* Deliberately modest. This is the one operation here that can allocate against the size of the
|
||
* PICTURE rather than against anything this service chose, so it is the one that can take a small
|
||
* machine down — and taking a machine down is a worse outcome than being slow, every time, because
|
||
* the other things on it did not ask to be part of an image conversion.
|
||
*
|
||
* 256 MB comfortably holds the working set of a phone photograph. Past it nothing fails: ImageMagick
|
||
* memory-maps and then falls back to disk, so a genuinely enormous image is slower rather than fatal.
|
||
*
|
||
* Raise it on a machine with headroom. See sharp/decodable.ts for how it is applied.
|
||
*/
|
||
export const CONVERT_MEMORY_MB = num("AV_CLUB_CONVERT_MEMORY_MB", 256, 32);
|
||
|
||
/** A crafted file can make an encoder run effectively forever. The process is killed at this point. */
|
||
export const JOB_TIMEOUT_MS = num("AV_CLUB_JOB_TIMEOUT_MS", 120_000, 1_000);
|
||
|
||
/**
|
||
* Decompression bombs: a 1 KB PNG can expand to tens of gigabytes. This is sharp's own
|
||
* `limitInputPixels`, defaulted to its documented default (0x3FFF × 0x3FFF) rather than to `false`.
|
||
*/
|
||
export const MAX_INPUT_PIXELS = num("AV_CLUB_MAX_INPUT_PIXELS", 0x3fff * 0x3fff, 1);
|
||
|
||
/** `resize(50000, 50000)` is an allocation request, not an image. */
|
||
export const MAX_OUTPUT_DIMENSION = num("AV_CLUB_MAX_OUTPUT_DIMENSION", 16_383, 1);
|
||
|
||
/** ffmpeg will take every core it is given. Two callers should not be able to take the box down. */
|
||
export const MAX_CONCURRENT_JOBS = num("AV_CLUB_MAX_CONCURRENT_JOBS", 2, 1);
|
||
|
||
/**
|
||
* The bearer token every request must carry.
|
||
*
|
||
* A FUNCTION, not a constant evaluated at import. The limits above are plain data that anything may
|
||
* read — the doc generator imports the sharp vocabulary, which imports MAX_OUTPUT_DIMENSION from
|
||
* here — and a module-scope throw meant that generating documentation required a production secret.
|
||
*
|
||
* The check itself still happens at boot: server.ts calls this before it listens, so a missing token
|
||
* is a refusal to start rather than a service that quietly accepts everyone. There is no
|
||
* "unauthenticated by default" mode; this process runs decoders over untrusted bytes.
|
||
*/
|
||
export function requireToken(): string {
|
||
const value = process.env.AV_CLUB_TOKEN;
|
||
if (!value || value.length < 16) {
|
||
throw new Error(
|
||
"AV_CLUB_TOKEN must be set and at least 16 characters. Generate one with: openssl rand -hex 32",
|
||
);
|
||
}
|
||
return value;
|
||
}
|