fix(upload): prevent false 422s and EMFILE crash during bulk uploads

Four related issues made uploading 271+ activities unreliable:

1. merge_all/write_combined_feed were inside the extraction try/except —
   any merge race returned 422 even though the file was on disk, causing
   the mobile app to permanently mark the upload as failed.  Fixed by
   moving them to a separate best-effort try/except after the extraction
   block.  Switch to merge_one (single-activity symlink) instead of
   merge_all (full rebuild) so each upload is O(1) FS ops, not O(N).

2. The dev watcher fired merge_all for every activity .json write AND the
   upload endpoint also ran merge_all — O(N²) symlink operations during
   bulk uploads.  Watcher now skips activities/*.json changes (upload
   endpoint handles those directly).

3. Vite/Chokidar followed the public/data symlink and opened a handle per
   activity file; constant merge rebuilds exhausted file descriptors and
   crashed the Astro dev server.  Fixed with watch.ignored on public/data.

4. _write_year_shards and write_combined_feed used f.unlink() without
   missing_ok=True — concurrent callers racing the same file threw
   FileNotFoundError which propagated as a false extraction failure.
This commit is contained in:
Davide Scaini
2026-04-27 14:33:05 +02:00
parent 7a65ed2078
commit be772bd3df
4 changed files with 34 additions and 9 deletions
+5
View File
@@ -141,6 +141,11 @@ def _watch_data(data: Path) -> None:
continue continue
for prefix, user_dir in prefix_to_user.items(): for prefix, user_dir in prefix_to_user.items():
if path.startswith(prefix): if path.startswith(prefix):
# Skip new-activity .json writes in activities/ — the upload
# endpoint calls merge_one inline, so firing merge_all here
# too would cause O(N²) full rebuilds during bulk uploads.
if path.startswith(str(user_dir / "activities")) and path.endswith(".json"):
break
affected.add(user_dir) affected.add(user_dir)
break break
+7 -2
View File
@@ -98,6 +98,11 @@ def merge_one(data_dir: Path, activity_id: str) -> None:
Use merge_all() for bulk operations (first run, Strava sync, etc.). Use merge_all() for bulk operations (first run, Strava sync, etc.).
""" """
with _merge_lock(data_dir):
_merge_one_locked(data_dir, activity_id)
def _merge_one_locked(data_dir: Path, activity_id: str) -> None:
edits_dir = data_dir / "edits" edits_dir = data_dir / "edits"
acts_dir = data_dir / "activities" acts_dir = data_dir / "activities"
merged_dir = data_dir / "_merged" merged_dir = data_dir / "_merged"
@@ -311,7 +316,7 @@ def _write_year_shards(merged_dir: Path, activities: list[dict], index_meta: dic
# Remove stale year shard files from previous runs # Remove stale year shard files from previous runs
for f in merged_dir.glob("index-*.json"): for f in merged_dir.glob("index-*.json"):
f.unlink() f.unlink(missing_ok=True)
by_year: dict[str, list[dict]] = defaultdict(list) by_year: dict[str, list[dict]] = defaultdict(list)
for a in activities: for a in activities:
@@ -398,7 +403,7 @@ def write_combined_feed(data_dir: Path) -> int:
# Remove stale feed pages # Remove stale feed pages
for f in data_dir.glob("feed*.json"): for f in data_dir.glob("feed*.json"):
f.unlink() f.unlink(missing_ok=True)
if not all_activities: if not all_activities:
return 0 return 0
+16 -7
View File
@@ -596,9 +596,12 @@ async def upload_bas_activity(
_upsert_index_summary(user_dir, activity_id, activity, geojson_body) _upsert_index_summary(user_dir, activity_id, activity, geojson_body)
from bincio.render.merge import merge_all, write_combined_feed try:
merge_all(user_dir) from bincio.render.merge import merge_one, write_combined_feed
write_combined_feed(_get_data_dir()) merge_one(user_dir, activity_id)
write_combined_feed(_get_data_dir())
except Exception as exc:
log.warning("upload/bas[%s]: merge/feed failed (non-fatal): %s", user.handle, exc)
log.info("upload/bas[%s]: imported %s", user.handle, activity_id) log.info("upload/bas[%s]: imported %s", user.handle, activity_id)
return JSONResponse({"ok": True, "id": activity_id, "status": "imported"}) return JSONResponse({"ok": True, "id": activity_id, "status": "imported"})
@@ -688,10 +691,6 @@ async def upload_raw_activity(
_upsert_index_summary(user_dir, act_id, detail, geojson) _upsert_index_summary(user_dir, act_id, detail, geojson)
from bincio.render.merge import merge_all, write_combined_feed
merge_all(user_dir)
write_combined_feed(_get_data_dir())
except Exception as exc: except Exception as exc:
log.warning("upload/raw[%s]: extraction failed: %s", user.handle, exc) log.warning("upload/raw[%s]: extraction failed: %s", user.handle, exc)
raise HTTPException(422, f"Could not extract activity: {exc}") from exc raise HTTPException(422, f"Could not extract activity: {exc}") from exc
@@ -699,6 +698,16 @@ async def upload_raw_activity(
tmp_in.unlink(missing_ok=True) tmp_in.unlink(missing_ok=True)
shutil.rmtree(tmp_out, ignore_errors=True) shutil.rmtree(tmp_out, ignore_errors=True)
# Merge and update feed — best effort; a race or transient FS error here must
# not turn a successful extraction into a 422 (the file is on disk; the mobile
# would retry indefinitely and the activity would never be marked synced).
try:
from bincio.render.merge import merge_one, write_combined_feed
merge_one(user_dir, act_id)
write_combined_feed(_get_data_dir())
except Exception as exc:
log.warning("upload/raw[%s]: merge/feed failed (non-fatal): %s", user.handle, exc)
log.info("upload/raw[%s]: imported %s", user.handle, act_id) log.info("upload/raw[%s]: imported %s", user.handle, act_id)
return JSONResponse({ return JSONResponse({
"ok": True, "ok": True,
+6
View File
@@ -39,6 +39,12 @@ export default defineConfig({
// Proxy /api/* to bincio serve/edit so cookies work same-origin in dev. // Proxy /api/* to bincio serve/edit so cookies work same-origin in dev.
// In production nginx handles this — same pattern, no code change needed. // In production nginx handles this — same pattern, no code change needed.
server: { server: {
watch: {
// public/data is a symlink to the live data dir; Chokidar follows it and
// opens a handle per file, which causes EMFILE during bulk uploads as
// activity count grows. Data files don't need HMR — exclude them.
ignored: ['**/public/data/**'],
},
proxy: { proxy: {
// Both /api/upload and /api/upload/strava-zip return SSE streams in response // Both /api/upload and /api/upload/strava-zip return SSE streams in response
// to POST requests. Vite's default proxy buffers the full body before forwarding, // to POST requests. Vite's default proxy buffers the full body before forwarding,