src/video/actors.rs

@@ -635,6 +635,95 @@ impl Handler<GeneratePlaylistMessage> for PlaylistGenerator {
 
             let nvenc = !use_copy && crate::video::ffmpeg::is_nvenc_available().await;
 
+            // Default timeout is generous — a long 4K transcode on CPU can take hours.
+            let timeout_secs = std::env::var("HLS_TIMEOUT_SECONDS")
+                .ok()
+                .and_then(|v| v.parse::<u64>().ok())
+                .unwrap_or(7200);
+
+            let mut ffmpeg_result = run_ffmpeg(
+                hls_command(&video_file, &playlist_tmp, &segment_pattern, use_copy, nvenc),
+                timeout_secs,
+            )
+            .await;
+
+            // A GPU that's busy right now (#215: llama-swap holding the 3090)
+            // says nothing about the file. Encode on the CPU instead of
+            // writing a sentinel that would skip this video for good.
+            if nvenc
+                && matches!(&ffmpeg_result, Ok(out) if !out.status.success()
+                    && crate::video::ffmpeg::is_gpu_init_failure(&String::from_utf8_lossy(&out.stderr)))
+            {
+                warn!(
+                    "GPU unavailable for {}, retrying the playlist on the CPU encoder",
+                    video_file
+                );
+                span.add_event("GPU unavailable, retrying on CPU", vec![]);
+                cleanup_partial_hls(&hash_dir).await;
+                ffmpeg_result = run_ffmpeg(
+                    hls_command(&video_file, &playlist_tmp, &segment_pattern, use_copy, false),
+                    timeout_secs,
+                )
+                .await;
+            }
+
+            drop(permit);
+
+            let success = matches!(&ffmpeg_result, Ok(out) if out.status.success());
+
+            if success {
+                if let Err(e) = tokio::fs::rename(&playlist_tmp, &playlist_path).await {
+                    error!(
+                        "ffmpeg succeeded but rename {} -> {} failed: {}",
+                        playlist_tmp, playlist_file, e
+                    );
+                    cleanup_partial_hls(&hash_dir).await;
+                    span.set_status(Status::error(format!("rename failed: {}", e)));
+                    return Err(e);
+                }
+                debug!("Playlist complete: {}", playlist_file);
+                span.set_status(Status::Ok);
+                Ok(())
+            } else {
+                let detail = match &ffmpeg_result {
+                    Ok(out) => format!(
+                        "exit {}: {}",
+                        out.status,
+                        String::from_utf8_lossy(&out.stderr).trim()
+                    ),
+                    Err(e) => format!("ffmpeg failed: {}", e),
+                };
+                error!("ffmpeg failed for {}: {}", video_file, detail);
+                cleanup_partial_hls(&hash_dir).await;
+                if let Err(se) = tokio::fs::write(&sentinel_path, b"").await {
+                    warn!(
+                        "Failed to write playlist sentinel {}: {}",
+                        sentinel_path.display(),
+                        se
+                    );
+                } else {
+                    info!(
+                        "Wrote playlist sentinel {} so future scans skip {}",
+                        sentinel_path.display(),
+                        video_file
+                    );
+                }
+                span.set_status(Status::error(detail.clone()));
+                Err(std::io::Error::other(detail))
+            }
+        }
+        .instrument(fut_span))
+    }
+}
+
+/// The ffmpeg command that writes a video's HLS playlist and segments.
+fn hls_command(
+    video_file: &str,
+    playlist_tmp: &str,
+    segment_pattern: &str,
+    use_copy: bool,
+    nvenc: bool,
+) -> tokio::process::Command {
     let mut cmd = tokio::process::Command::new("ffmpeg");
     cmd.arg("-y");
     if nvenc {
@@ -645,7 +734,7 @@ impl Handler<GeneratePlaylistMessage> for PlaylistGenerator {
         // back to software decode for codecs NVDEC lacks (ProRes).
         cmd.arg("-hwaccel").arg("cuda");
     }
-            cmd.arg("-i").arg(&video_file);
+    cmd.arg("-i").arg(video_file);
 
     if use_copy {
         cmd.arg("-c:v").arg("copy");
@@ -701,21 +790,21 @@ impl Handler<GeneratePlaylistMessage> for PlaylistGenerator {
     // decoded without reference to any other — the matching guarantee
     // for the forced keyframes above.
     cmd.arg("-hls_flags").arg("independent_segments");
-            cmd.arg("-hls_segment_filename").arg(&segment_pattern);
-            cmd.arg(&playlist_tmp);
+    cmd.arg("-hls_segment_filename").arg(segment_pattern);
+    cmd.arg(playlist_tmp);
     cmd.stdout(Stdio::null());
     cmd.stderr(Stdio::piped());
     cmd.kill_on_drop(true);
+    cmd
+}
 
-            // Spawn + wait under a timeout so a hung ffmpeg (corrupt source,
-            // NFS stall, etc.) doesn't permanently hold a semaphore slot.
-            // Default is generous — a long 4K transcode on CPU can take hours.
-            let timeout_secs = std::env::var("HLS_TIMEOUT_SECONDS")
-                .ok()
-                .and_then(|v| v.parse::<u64>().ok())
-                .unwrap_or(7200);
-
-            let ffmpeg_result = match cmd.spawn() {
+/// Run ffmpeg under a timeout, so a hung run (corrupt source, NFS stall)
+/// doesn't hold a semaphore slot forever.
+async fn run_ffmpeg(
+    mut cmd: tokio::process::Command,
+    timeout_secs: u64,
+) -> Result<std::process::Output> {
+    match cmd.spawn() {
         Ok(child) => {
             match tokio::time::timeout(
                 std::time::Duration::from_secs(timeout_secs),
@@ -724,9 +813,7 @@ impl Handler<GeneratePlaylistMessage> for PlaylistGenerator {
             .await
             {
                 Ok(res) => res
-                            .inspect_err(|e| {
-                                error!("Failed to wait on ffmpeg child process: {}", e)
-                            })
+                    .inspect_err(|e| error!("Failed to wait on ffmpeg child process: {}", e))
                     .map_err(|e| std::io::Error::other(e.to_string())),
                 Err(_) => Err(std::io::Error::other(format!(
                     "ffmpeg exceeded {}s timeout",
@@ -738,54 +825,6 @@ impl Handler<GeneratePlaylistMessage> for PlaylistGenerator {
             error!("Failed to spawn ffmpeg: {}", e);
             Err(std::io::Error::other(e.to_string()))
         }
-            };
-
-            drop(permit);
-
-            let success = matches!(&ffmpeg_result, Ok(out) if out.status.success());
-
-            if success {
-                if let Err(e) = tokio::fs::rename(&playlist_tmp, &playlist_path).await {
-                    error!(
-                        "ffmpeg succeeded but rename {} -> {} failed: {}",
-                        playlist_tmp, playlist_file, e
-                    );
-                    cleanup_partial_hls(&hash_dir).await;
-                    span.set_status(Status::error(format!("rename failed: {}", e)));
-                    return Err(e);
-                }
-                debug!("Playlist complete: {}", playlist_file);
-                span.set_status(Status::Ok);
-                Ok(())
-            } else {
-                let detail = match &ffmpeg_result {
-                    Ok(out) => format!(
-                        "exit {}: {}",
-                        out.status,
-                        String::from_utf8_lossy(&out.stderr).trim()
-                    ),
-                    Err(e) => format!("ffmpeg failed: {}", e),
-                };
-                error!("ffmpeg failed for {}: {}", video_file, detail);
-                cleanup_partial_hls(&hash_dir).await;
-                if let Err(se) = tokio::fs::write(&sentinel_path, b"").await {
-                    warn!(
-                        "Failed to write playlist sentinel {}: {}",
-                        sentinel_path.display(),
-                        se
-                    );
-                } else {
-                    info!(
-                        "Wrote playlist sentinel {} so future scans skip {}",
-                        sentinel_path.display(),
-                        video_file
-                    );
-                }
-                span.set_status(Status::error(detail.clone()));
-                Err(std::io::Error::other(detail))
-            }
-        }
-        .instrument(fut_span))
     }
 }
 
@@ -958,7 +997,29 @@ impl Handler<GeneratePreviewClipMessage> for PreviewClipGenerator {
 
 #[cfg(test)]
 mod tests {
-    use super::{VideoStreamMeta, fits_hls_copy_limits, parse_ffprobe_rational};
+    use super::{VideoStreamMeta, fits_hls_copy_limits, hls_command, parse_ffprobe_rational};
+
+    fn args(nvenc: bool) -> Vec<String> {
+        hls_command("in.mp4", "out.m3u8.tmp", "seg_%03d.ts", false, nvenc)
+            .as_std()
+            .get_args()
+            .map(|a| a.to_string_lossy().into_owned())
+            .collect()
+    }
+
+    #[test]
+    fn cpu_fallback_command_never_touches_the_gpu() {
+        // #215: the retry after a GPU init failure must not ask for CUDA again.
+        let cpu = args(false);
+        assert!(!cpu.iter().any(|a| a == "-hwaccel" || a == "h264_nvenc"));
+        assert!(cpu.windows(2).any(|w| w[0] == "-c:v" && w[1] == "h264"));
+        let gpu = args(true);
+        assert!(gpu.windows(2).any(|w| w[0] == "-hwaccel" && w[1] == "cuda"));
+        assert!(
+            gpu.windows(2)
+                .any(|w| w[0] == "-c:v" && w[1] == "h264_nvenc")
+        );
+    }
 
     const LIMIT: u64 = 20_000_000;
 

src/video/ffmpeg.rs

@@ -36,6 +36,24 @@ pub async fn is_nvenc_available() -> bool {
     available
 }
 
+/// Whether ffmpeg's stderr says the GPU couldn't be used *right now*, as
+/// opposed to the input being bad (#215). `is_nvenc_available` only checks
+/// that ffmpeg was built with NVENC; on a GPU shared with llama-swap a resident
+/// model can leave no memory for a CUDA context, and ffmpeg then exits 187
+/// with `cuCtxCreate(...) failed -> CUDA_ERROR_OUT_OF_MEMORY`. Callers rerun
+/// such a failure on the CPU encoder instead of recording the file as broken.
+pub fn is_gpu_init_failure(stderr: &str) -> bool {
+    const MARKERS: &[&str] = &[
+        "CUDA_ERROR_",
+        "cuCtxCreate",
+        "OpenEncodeSessionEx failed",
+        "No capable devices found",
+        "Cannot load libcuda",
+        "Device creation failed",
+    ];
+    MARKERS.iter().any(|m| stderr.contains(m))
+}
+
 pub struct Ffmpeg;
 
 pub enum GifType {
@@ -271,25 +289,13 @@ fn parse_ffprobe_duration(stdout: &str) -> Option<f64> {
     None
 }
 
-/// Generate a preview clip from a video file.
-///
-/// Creates a ~10 second MP4 by extracting up to 10 equally-spaced 1-second segments
-/// at 480p with H.264 video and AAC audio. For short videos (<10s), uses fewer segments.
-/// For very short videos (<1s), transcodes the entire video.
-///
-/// Returns (duration_seconds, file_size_bytes) on success.
-pub async fn generate_preview_clip(input_file: &str, output_file: &str) -> Result<(f64, u64)> {
-    info!("Generating preview clip for: '{}'", input_file);
-    let start = Instant::now();
-
-    let duration = get_duration_seconds(input_file).await?;
-    let use_nvenc = is_nvenc_available().await;
-
-    // Create parent directories for output
-    if let Some(parent) = std::path::Path::new(output_file).parent() {
-        std::fs::create_dir_all(parent)?;
-    }
-
+/// The ffmpeg command for a preview clip, and the clip duration it produces.
+fn preview_clip_command(
+    input_file: &str,
+    output_file: &str,
+    duration: Option<f64>,
+    use_nvenc: bool,
+) -> (Command, f64) {
     let mut cmd = Command::new("ffmpeg");
     cmd.arg("-y");
 
@@ -350,7 +356,43 @@ pub async fn generate_preview_clip(input_file: &str, output_file: &str) -> Resul
     cmd.stdout(Stdio::null());
     cmd.stderr(Stdio::piped());
 
-    let output = cmd.output().await?;
+    (cmd, clip_duration)
+}
+
+/// Generate a preview clip from a video file.
+///
+/// Creates a ~10 second MP4 by extracting up to 10 equally-spaced 1-second segments
+/// at 480p with H.264 video and AAC audio. For short videos (<10s), uses fewer segments.
+/// For very short videos (<1s), transcodes the entire video.
+///
+/// Returns (duration_seconds, file_size_bytes) on success.
+pub async fn generate_preview_clip(input_file: &str, output_file: &str) -> Result<(f64, u64)> {
+    info!("Generating preview clip for: '{}'", input_file);
+    let start = Instant::now();
+
+    let duration = get_duration_seconds(input_file).await?;
+    let use_nvenc = is_nvenc_available().await;
+
+    // Create parent directories for output
+    if let Some(parent) = std::path::Path::new(output_file).parent() {
+        std::fs::create_dir_all(parent)?;
+    }
+
+    let (mut cmd, clip_duration) =
+        preview_clip_command(input_file, output_file, duration, use_nvenc);
+    let mut output = cmd.output().await?;
+
+    if use_nvenc
+        && !output.status.success()
+        && is_gpu_init_failure(&String::from_utf8_lossy(&output.stderr))
+    {
+        warn!(
+            "GPU unavailable for preview of '{}', retrying on the CPU encoder",
+            input_file
+        );
+        let (mut cmd, _) = preview_clip_command(input_file, output_file, duration, false);
+        output = cmd.output().await?;
+    }
 
     if !output.status.success() {
         let stderr = String::from_utf8_lossy(&output.stderr);
@@ -376,7 +418,28 @@ pub async fn generate_preview_clip(input_file: &str, output_file: &str) -> Resul
 
 #[cfg(test)]
 mod tests {
-    use super::parse_ffprobe_duration;
+    use super::{is_gpu_init_failure, parse_ffprobe_duration};
+
+    #[test]
+    fn gpu_busy_is_a_gpu_init_failure() {
+        // Verbatim from the e2e server with a llama-swap model resident (#215).
+        let oom = "[CUDA @ 0x5d1c2c0] cu->cuCtxCreate(&hwctx->cuda_ctx, desired_flags, hwctx->internal->cuda_device) failed -> CUDA_ERROR_OUT_OF_MEMORY: out of memory\nDevice creation failed: -542398533.";
+        assert!(is_gpu_init_failure(oom));
+        assert!(is_gpu_init_failure(
+            "[h264_nvenc @ 0x1] OpenEncodeSessionEx failed: out of memory (10): (no details)"
+        ));
+        assert!(is_gpu_init_failure(
+            "[h264_nvenc @ 0x1] No capable devices found"
+        ));
+    }
+
+    #[test]
+    fn bad_input_is_not_a_gpu_init_failure() {
+        assert!(!is_gpu_init_failure(
+            "[mov,mp4,m4a,3gp,3g2,mj2 @ 0x1] moov atom not found\nclip.mp4: Invalid data found when processing input"
+        ));
+        assert!(!is_gpu_init_failure("Conversion failed!"));
+    }
 
     #[test]
     fn empty_output_returns_none() {