feat(serve): debounced site rebuild — burst uploads trigger one build, not N
Replace per-upload Astro build threads with a single background worker (_site_rebuild_worker) that waits on an event, sleeps 60 s to let upload bursts settle, then runs one full build + rsync. 271 concurrent uploads now produce one build instead of 271 serialised builds, eliminating the OOM kill. --webroot is re-enabled; merge-only path still runs immediately per upload. Also: date filter row added to ActivityFeed.svelte (sport + date presets with dynamic year pills); deploy/vps gitignored for VPS config backups.
This commit is contained in:
Davide Scaini
+77
-68
@@ -204,8 +204,8 @@ app = FastAPI(title="BincioActivity Serve")
|
||||
|
||||
|
||||
@app.on_event("startup")
|
||||
async def _cleanup_orphaned_tmp_zips() -> None:
|
||||
"""Remove tmp*.zip files left in user data dirs by the pre-fix upload handler."""
|
||||
async def _on_startup() -> None:
|
||||
"""Startup tasks: clean orphaned tmp zips; launch site-rebuild worker if --webroot set."""
|
||||
import glob as _glob
|
||||
data_dir = _get_data_dir()
|
||||
for p in _glob.glob(str(data_dir / "*" / "tmp*.zip")):
|
||||
@@ -213,6 +213,8 @@ async def _cleanup_orphaned_tmp_zips() -> None:
|
||||
Path(p).unlink()
|
||||
except Exception:
|
||||
pass
|
||||
if webroot is not None:
|
||||
threading.Thread(target=_site_rebuild_worker, daemon=True, name="site-rebuild").start()
|
||||
|
||||
|
||||
app.add_middleware(GZipMiddleware, minimum_size=1024)
|
||||
@@ -329,20 +331,65 @@ def _unique_image_name(directory: Path, filename: str) -> str:
|
||||
|
||||
# ── Post-write rebuild ────────────────────────────────────────────────────────
|
||||
|
||||
# Serialises concurrent rebuilds — only one full build runs at a time.
|
||||
# A second upload that arrives while a build is in progress will queue and
|
||||
# run after the first finishes, picking up all data written in between.
|
||||
# Serialises per-user merge subprocesses — concurrent merge_all runs on the
|
||||
# same user dir would corrupt _merged/activities/.
|
||||
_rebuild_lock = threading.Lock()
|
||||
|
||||
# Signals the site-rebuild worker that at least one merge has completed.
|
||||
# Using an Event as a boolean flag: set() by any merge, cleared by the worker.
|
||||
_site_rebuild_event = threading.Event()
|
||||
|
||||
|
||||
def _site_rebuild_worker() -> None:
|
||||
"""Single background thread: debounced Astro build + rsync after uploads.
|
||||
|
||||
Waits for _site_rebuild_event, sleeps 60 s to let upload bursts settle,
|
||||
then runs one full build. 271 concurrent uploads → 1 build, not 271.
|
||||
Uploads that arrive during the build set the event again, so a follow-up
|
||||
build starts after the current one finishes.
|
||||
"""
|
||||
_webroot = str(webroot)
|
||||
_data_dir = str(data_dir)
|
||||
_site_dir = str(site_dir)
|
||||
uv = shutil.which("uv") or str(Path.home() / ".local" / "bin" / "uv")
|
||||
while True:
|
||||
_site_rebuild_event.wait()
|
||||
_site_rebuild_event.clear()
|
||||
time.sleep(60) # collect burst uploads
|
||||
_site_rebuild_event.clear() # discard signals from the sleep window
|
||||
log.info("site-rebuild: starting full build + rsync to %s", _webroot)
|
||||
try:
|
||||
result = subprocess.run(
|
||||
[uv, "run", "bincio", "render",
|
||||
"--data-dir", _data_dir,
|
||||
"--site-dir", _site_dir],
|
||||
capture_output=True,
|
||||
text=True,
|
||||
)
|
||||
if result.returncode != 0:
|
||||
log.error("site-rebuild: build failed (rc=%d):\n%s\n%s",
|
||||
result.returncode, result.stdout, result.stderr)
|
||||
continue
|
||||
dist_data = Path(_site_dir) / "dist" / "data"
|
||||
if dist_data.exists():
|
||||
shutil.rmtree(dist_data)
|
||||
rsync = subprocess.run(
|
||||
["rsync", "-a", "--delete", "--exclude=data/",
|
||||
f"{_site_dir}/dist/", _webroot + "/"],
|
||||
capture_output=True,
|
||||
text=True,
|
||||
)
|
||||
if rsync.returncode != 0:
|
||||
log.error("site-rebuild: rsync failed (rc=%d):\n%s\n%s",
|
||||
rsync.returncode, rsync.stdout, rsync.stderr)
|
||||
else:
|
||||
log.info("site-rebuild: done")
|
||||
except Exception:
|
||||
log.exception("site-rebuild: unexpected error")
|
||||
|
||||
|
||||
def _trigger_rebuild(handle: str) -> None:
|
||||
"""Asynchronously re-merge and optionally rebuild + rsync the site.
|
||||
|
||||
- Without --webroot: fast path — merges sidecars + rewrites root manifest
|
||||
(~1 s). New activity pages require the nginx try_files fallback to work.
|
||||
- With --webroot: full Astro build + rsync to the nginx webroot (~30–60 s,
|
||||
serialised). New activity pages are immediately accessible.
|
||||
"""
|
||||
"""Merge sidecars for handle asynchronously; signal the site-rebuild worker."""
|
||||
if site_dir is None:
|
||||
return
|
||||
if not _VALID_HANDLE.match(handle):
|
||||
@@ -351,66 +398,28 @@ def _trigger_rebuild(handle: str) -> None:
|
||||
uv = shutil.which("uv") or str(Path.home() / ".local" / "bin" / "uv")
|
||||
_data_dir = str(data_dir)
|
||||
_site_dir = str(site_dir)
|
||||
_webroot = str(webroot) if webroot else None
|
||||
_handle = handle
|
||||
|
||||
def _run() -> None:
|
||||
try:
|
||||
if _webroot is None:
|
||||
# Fast: only update data, skip Astro build.
|
||||
# Serialised with the same lock: merge_all wipes and recreates
|
||||
# _merged/activities/ — concurrent runs would corrupt each other.
|
||||
log.info("rebuild[%s]: merge-only (no webroot)", _handle)
|
||||
with _rebuild_lock:
|
||||
result = subprocess.run(
|
||||
[uv, "run", "bincio", "render",
|
||||
"--data-dir", _data_dir,
|
||||
"--site-dir", _site_dir,
|
||||
"--handle", _handle,
|
||||
"--no-build"],
|
||||
capture_output=True,
|
||||
text=True,
|
||||
)
|
||||
if result.returncode != 0:
|
||||
log.error("rebuild[%s]: merge failed (rc=%d):\n%s\n%s",
|
||||
_handle, result.returncode, result.stdout, result.stderr)
|
||||
else:
|
||||
log.info("rebuild[%s]: merge done", _handle)
|
||||
else:
|
||||
# Full build + rsync — serialised so concurrent uploads don't race
|
||||
log.info("rebuild[%s]: full build + rsync to %s", _handle, _webroot)
|
||||
with _rebuild_lock:
|
||||
result = subprocess.run(
|
||||
[uv, "run", "bincio", "render",
|
||||
"--data-dir", _data_dir,
|
||||
"--site-dir", _site_dir,
|
||||
"--handle", _handle],
|
||||
capture_output=True,
|
||||
text=True,
|
||||
)
|
||||
if result.returncode != 0:
|
||||
log.error("rebuild[%s]: build failed (rc=%d):\n%s\n%s",
|
||||
_handle, result.returncode, result.stdout, result.stderr)
|
||||
else:
|
||||
log.info("rebuild[%s]: build done, rsyncing", _handle)
|
||||
# Prune dist/data/ before rsync: Astro resolves the
|
||||
# public/data symlink and copies all activity JSON into
|
||||
# dist/, but nginx already serves /data/ directly from
|
||||
# the live data dir — rsyncing it would duplicate GBs.
|
||||
dist_data = Path(_site_dir) / "dist" / "data"
|
||||
if dist_data.exists():
|
||||
shutil.rmtree(dist_data)
|
||||
rsync = subprocess.run(
|
||||
["rsync", "-a", "--delete", "--exclude=data/",
|
||||
f"{_site_dir}/dist/", _webroot + "/"],
|
||||
capture_output=True,
|
||||
text=True,
|
||||
)
|
||||
if rsync.returncode != 0:
|
||||
log.error("rebuild[%s]: rsync failed (rc=%d):\n%s\n%s",
|
||||
_handle, rsync.returncode, rsync.stdout, rsync.stderr)
|
||||
else:
|
||||
log.info("rebuild[%s]: rsync done", _handle)
|
||||
log.info("rebuild[%s]: merge-only", _handle)
|
||||
with _rebuild_lock:
|
||||
result = subprocess.run(
|
||||
[uv, "run", "bincio", "render",
|
||||
"--data-dir", _data_dir,
|
||||
"--site-dir", _site_dir,
|
||||
"--handle", _handle,
|
||||
"--no-build"],
|
||||
capture_output=True,
|
||||
text=True,
|
||||
)
|
||||
if result.returncode != 0:
|
||||
log.error("rebuild[%s]: merge failed (rc=%d):\n%s\n%s",
|
||||
_handle, result.returncode, result.stdout, result.stderr)
|
||||
else:
|
||||
log.info("rebuild[%s]: merge done", _handle)
|
||||
if webroot is not None:
|
||||
_site_rebuild_event.set()
|
||||
except Exception:
|
||||
log.exception("rebuild[%s]: unexpected error", _handle)
|
||||
|
||||
|
||||
Reference in New Issue
Block a user