Skip to content

fix(render-farm): clean up failed video outputs #2898

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
richiemcilroy merged 2 commits into from
Oct 22, 2026
Merged

fix(render-farm): clean up failed video outputs #2898

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
a42705b
fix(render-farm): clean up failed video outputs
yashranaway Oct 22, 2026
bb54786
Merge branch 'main' into fix/render-farm-worker
richiemcilroy Oct 22, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
66 changes: 64 additions & 2 deletions apps/render-farm/src/worker.test.ts
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2,7 +2,12 @@ import { describe, expect, test } from "bun:test";
import { randomUUID } from "node:crypto";
import { readFileSync } from "node:fs";
import { join } from "node:path";
import { MIN_PART, type TranscodeTask, type WorkItem } from "./protocol";
import {
MIN_PART,
type TranscodeTask,
type VideoTask,
type WorkItem,
} from "./protocol";
import * as stitch from "./stitch";
import * as transcode from "./transcode";

Expand All @@ -26,6 +31,8 @@ function harness(
randomUUID,
join,
...transcode,
processCpuSeconds: async () => 0,
sampleThreads: async () => ({}),
availableParallelism: () => 1,
hostname: () => "worker-test",
mkdirSync: () => {},
Expand All @@ -34,6 +41,13 @@ function harness(
s3ConfigFromEnv: () => ({}),
...stitch,
MIN_PART,
ProjectCache: class {
root = "/work/job";
async materialize() {
return { bytes: 0, ms: 0, requests: 0 };
}
close() {}
},
S3: class {
async presignFresh() {
await options.presignGate;
Expand Down Expand Up @@ -97,8 +111,18 @@ function harness(
const compiled = new Bun.Transpiler({ loader: "ts" }).transformSync(source);
const worker = new Function(
...Object.keys(deps),
`${compiled}\nreturn {runTranscode, cancel, busy, progress, transcoders, engineEnv, uploadLayout};`,
`${compiled}\nreturn {runVideo, runTranscode, cancel, busy, progress, transcoders, engineEnv, uploadLayout};`,
)(...Object.values(deps)) as {
runVideo: (
task: VideoTask,
engine: {
pid: number;
onEvent: ((event: Record<string, unknown>) => void) | null;
onProgress: ((frames: number) => void) | null;
request: () => Promise<never>;
},
queuedMs: number,
) => Promise<unknown>;
runTranscode: (task: TranscodeTask, slot: number) => Promise<number>;
cancel: (taskIds: string[]) => void;
busy: Map<number, WorkItem>;
Expand Down Expand Up @@ -227,3 +251,41 @@ describe("chunk upload", () => {
});
});
});

test("a failed video render removes its temporary output", async () => {
const worker = harness();
const task: VideoTask = {
kind: "video",
taskId: "job:v0",
jobId: "job",
chunk: 0,
fps: 30,
resolution: [1920, 1080],
bpp: 0.1,
frames: [0, 30],
threads: 1,
files: [],
upload: {
key: "out/job.mp4",
uploadId: "upload",
firstPart: 2,
partLimit: 3,
partTarget: MIN_PART,
stashKey: "stash/job/0",
},
audio: null,
hls: null,
};
const engine = {
pid: 1,
onEvent: null,
onProgress: null,
request: () => Promise.reject(new Error("render failed")),
};
await expect(worker.runVideo(task, engine, 0)).rejects.toThrow(
"render failed",
);
expect(worker.removed).toHaveLength(1);
expect(worker.removed[0]).toStartWith("/work/job/v0-");
expect(worker.removed[0]).toEndWith(".h264");
});
214 changes: 110 additions & 104 deletions apps/render-farm/src/worker.ts
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -505,110 +505,116 @@ async function runVideo(
cache.root,
`v${task.chunk}-${randomUUID().slice(0, 6)}.h264`,
);
const engineStarted = performance.now();
const slot = engines.indexOf(engine);
const entry = {
taskId: task.taskId,
kind: "video" as const,
frames: 0,
total: task.frames[1] - task.frames[0],
startedAt: Date.now(),
lastProgressAt: Date.now(),
};
progress.set(slot, entry);
const stream = task.hls
? new SegmentStream({ ...task, hls: task.hls }, out, engineStarted)
: null;
engine.onEvent = stream ? (event) => stream.onEvent(event) : null;
// Ask for the next task once this one is ~PREFETCH_LEAD_MS from done, so
// its sources download while this one finishes instead of after.
engine.onProgress = (frames) => {
entry.frames = frames;
entry.lastProgressAt = Date.now();
const elapsed = performance.now() - engineStarted;
const remainingMs =
((entry.total - frames) * elapsed) / Math.max(1, frames);
if (frames > 0 && remainingMs <= PREFETCH_LEAD_MS) nearEnd();
};
const request = engine.request<{
width: number;
height: number;
sizes: number[];
keyframes: number[];
extradata: string;
bytes: number;
timings: Record<string, number>;
}>("video", {
project: cache.root,
fps: task.fps,
resolution: task.resolution,
bpp: task.bpp,
frames: task.frames,
threads: task.threads,
out,
});
const threadsPromise: Promise<Record<string, number>> =
process.env.RF_PROFILE_THREADS === "1"
? sampleThreads(engine.pid, request)
: Promise.resolve({});
const result = await request.finally(() => {
engine.onProgress = null;
engine.onEvent = null;
progress.delete(slot);
});
nearEnd();
const segmentsDone = stream?.finish(result.sizes, result.extradata);
segmentsDone?.catch(() => {});
const engineMs = performance.now() - engineStarted;
const threads = await threadsPromise;
const waitStarted = performance.now();
const audio = await audioPromise;
const audioWaitMs = performance.now() - waitStarted;
const uploadStarted = performance.now();
const plan = layout(task, out, result.sizes, audio);
const [{ parts, stash }, segments] = await Promise.all([
uploadLayout(task, plan.segments, plan.bytes),
segmentsDone ?? Promise.resolve(0),
]);
const uploadMs = performance.now() - uploadStarted;
rmSync(out, { force: true });
const timings: TaskTimings = {
queuedMs,
fetch: fetchStats,
engine: {
...result.timings,
...Object.fromEntries(
Object.entries(threads).map(([name, value]) => [
`thread:${name}`,
value,
]),
),
...(stream
? { segments, firstSegmentMs: Math.round(stream.firstUploadedMs ?? -1) }
: {}),
},
engineMs,
audioWaitMs,
uploadMs,
totalMs: performance.now() - started,
cpuSeconds: (await processCpuSeconds(engine.pid)) - cpuBefore,
};
return {
taskId: task.taskId,
worker: WORKER_ID,
sizes: result.sizes,
keyframes: result.keyframes,
extradata: result.extradata,
width: result.width,
height: result.height,
videoRuns: plan.videoRuns,
audioRuns: plan.audioRuns,
firstPart: task.upload.firstPart,
stash,
parts,
bytes: plan.bytes,
timings,
};
try {
const engineStarted = performance.now();
const slot = engines.indexOf(engine);
const entry = {
taskId: task.taskId,
kind: "video" as const,
frames: 0,
total: task.frames[1] - task.frames[0],
startedAt: Date.now(),
lastProgressAt: Date.now(),
};
progress.set(slot, entry);
const stream = task.hls
? new SegmentStream({ ...task, hls: task.hls }, out, engineStarted)
: null;
engine.onEvent = stream ? (event) => stream.onEvent(event) : null;
// Ask for the next task once this one is ~PREFETCH_LEAD_MS from done, so
// its sources download while this one finishes instead of after.
engine.onProgress = (frames) => {
entry.frames = frames;
entry.lastProgressAt = Date.now();
const elapsed = performance.now() - engineStarted;
const remainingMs =
((entry.total - frames) * elapsed) / Math.max(1, frames);
if (frames > 0 && remainingMs <= PREFETCH_LEAD_MS) nearEnd();
};
const request = engine.request<{
width: number;
height: number;
sizes: number[];
keyframes: number[];
extradata: string;
bytes: number;
timings: Record<string, number>;
}>("video", {
project: cache.root,
fps: task.fps,
resolution: task.resolution,
bpp: task.bpp,
frames: task.frames,
threads: task.threads,
out,
});
const threadsPromise: Promise<Record<string, number>> =
process.env.RF_PROFILE_THREADS === "1"
? sampleThreads(engine.pid, request)
: Promise.resolve({});
const result = await request.finally(() => {
engine.onProgress = null;
engine.onEvent = null;
progress.delete(slot);
});
nearEnd();
const segmentsDone = stream?.finish(result.sizes, result.extradata);
segmentsDone?.catch(() => {});
const engineMs = performance.now() - engineStarted;
const threads = await threadsPromise;
const waitStarted = performance.now();
const audio = await audioPromise;
const audioWaitMs = performance.now() - waitStarted;
const uploadStarted = performance.now();
const plan = layout(task, out, result.sizes, audio);
const [{ parts, stash }, segments] = await Promise.all([
uploadLayout(task, plan.segments, plan.bytes),
segmentsDone ?? Promise.resolve(0),
]);
const uploadMs = performance.now() - uploadStarted;
const timings: TaskTimings = {
queuedMs,
fetch: fetchStats,
engine: {
...result.timings,
...Object.fromEntries(
Object.entries(threads).map(([name, value]) => [
`thread:${name}`,
value,
]),
),
...(stream
? {
segments,
firstSegmentMs: Math.round(stream.firstUploadedMs ?? -1),
}
: {}),
},
engineMs,
audioWaitMs,
uploadMs,
totalMs: performance.now() - started,
cpuSeconds: (await processCpuSeconds(engine.pid)) - cpuBefore,
};
return {
taskId: task.taskId,
worker: WORKER_ID,
sizes: result.sizes,
keyframes: result.keyframes,
extradata: result.extradata,
width: result.width,
height: result.height,
videoRuns: plan.videoRuns,
audioRuns: plan.audioRuns,
firstPart: task.upload.firstPart,
stash,
parts,
bytes: plan.bytes,
timings,
};
} finally {
rmSync(out, { force: true });
}
}

async function runAudio(task: AudioTask, engine: Engine, queuedMs: number) {
Expand Down
Loading