Stream live worker progress via async build dispatch

The agent's /upload and /checkout_git were synchronous, blocking until a
build finished, so the controller could not relay live per-step progress.
Add ?async=1 support: the agent saves/clones the source, starts the build in
a background thread (holding build_lock), and returns 202 + {"id": ...}
immediately. The controller dispatches async, then polls /progress/<job_id>,
appending each step/log line to the job log and SSE stream, and treats the
build as done when progress clears (404) and the worker reports ready.
This commit is contained in:
Brett Williams
2026-08-30 21:59:46 -05:00
parent 0e89919342
commit b5b1cd740b
3 changed files with 102 additions and 32 deletions
+14 -13
View File
@@ -44,8 +44,8 @@ trusted-LAN posture as today). The worker endpoint contract used by the controll
| Worker endpoint | Method | Purpose | | Worker endpoint | Method | Purpose |
|---|---|---| |---|---|---|
| `/status` | GET | readiness, os, cpu, agent_version, hostname, ip | | `/status` | GET | readiness, os, cpu, agent_version, hostname, ip |
| `/upload` | POST | multipart `file` zip → starts build, returns job JSON | | `/upload` | POST | multipart `file` zip → starts build, returns job JSON. `?async=1` returns `202` + `{"id": ...}` immediately, building in background |
| `/checkout_git` | POST | JSON `{"repo_url": ...}` → starts build | | `/checkout_git` | POST | JSON `{"repo_url": ...}` → starts build. `?async=1` behaves like `/upload?async=1` |
| `/download/<job_id>` | GET | zip of `dist/` for a built job | | `/download/<job_id>` | GET | zip of `dist/` for a built job |
| `/delete_cache` | GET | clear cached jobs | | `/delete_cache` | GET | clear cached jobs |
| `/healthz` | GET | liveness (already added) | | `/healthz` | GET | liveness (already added) |
@@ -108,17 +108,18 @@ A single background thread owns all dispatch work.
- exclude the local/controller LXC if present, to avoid self-builds - exclude the local/controller LXC if present, to avoid self-builds
3. Choose first match (optionally: prefer workers with more free disk — v1: first match). 3. Choose first match (optionally: prefer workers with more free disk — v1: first match).
4. Mark job `dispatching`; set `worker_host`. 4. Mark job `dispatching`; set `worker_host`.
5. Send to worker: 5. Send to worker (**async**): `POST /upload?async=1` (with the stored file) or
- upload → `POST /upload` with the stored file `POST /checkout_git?async=1` (with `{repo_url}`). The agent clones/saves the source, then
- git → `POST /checkout_git` with `{repo_url}` starts the build in a background thread (holding its `build_lock`) and returns `202`
- on success, read the worker job JSON → store `worker_job_id`, `worker_url`, set `building` immediately with `{"id": <worker_job_id>}`. On hard failure, mark `failed` (§ failure handling
- on hard failure, mark `failed` (§ failure handling below) below). Store `worker_job_id`, `worker_url`, set `building`.
6. **Progress relay** — while the dispatch call is in flight (the worker is building), poll 6. **Progress relay** — while the build runs, poll `GET /progress/<worker_job_id>` on an interval,
`GET /progress/<worker_job_id>` on an interval and append each snapshot to the job log + appending each `last_log_line`/step snapshot to the job log and pushing to connected SSE
push it to connected SSE clients. This gives live per-step progress in the UI, enabled by clients → live per-step progress in the UI. The build is complete when `/progress` returns
the small agent addition (§1). `404` (progress cleared) and the worker's `/status` reports `ready`. (This relies on the small
7. On success (dispatch call returns): compute download URL `worker_url + /download/<worker_job_id>`, agent additions in §1.)
fetch the zip, save to `builds/<job_id>/`, record artifact in the job row, set `done`. 7. On completion: compute download URL `worker_url + /download/<worker_job_id>`, fetch the zip,
save to `builds/<job_id>/`, record artifact in the job row, set `done`.
**Concurrency/limits:** scheduler processes one job at a time (workers can already build only one **Concurrency/limits:** scheduler processes one job at a time (workers can already build only one
at a time — `409`). Multiple queued jobs simply wait. This makes NIC/disk behavior predictable at a time — `409`). Multiple queued jobs simply wait. This makes NIC/disk behavior predictable
+33 -3
View File
@@ -220,6 +220,7 @@ def checkout_project():
if not build_lock.acquire(blocking=False): if not build_lock.acquire(blocking=False):
return jsonify({'error': f"Another build is already running: {system_status['running_job']}"}), 409 return jsonify({'error': f"Another build is already running: {system_status['running_job']}"}), 409
async_mode = _is_async_request()
repo_url = request.json['repo_url'] repo_url = request.json['repo_url']
print(f"\n========== Checking Out Git Project ==========") print(f"\n========== Checking Out Git Project ==========")
@@ -233,6 +234,9 @@ def checkout_project():
system_status['status'] = "cloning_repo" system_status['status'] = "cloning_repo"
subprocess.check_call(['git', 'clone', repo_url, repo_dir]) subprocess.check_call(['git', 'clone', repo_url, repo_dir])
system_status['status'] = "ready" system_status['status'] = "ready"
if async_mode:
_run_build_background(job_id, repo_dir, start_time)
return jsonify({"id": job_id, "async": True, "message": "Build started"}), 202
return install_and_build(repo_dir, job_id, start_time) return install_and_build(repo_dir, job_id, start_time)
except Exception as e: except Exception as e:
print(f"Error processing checkout: {e}") print(f"Error processing checkout: {e}")
@@ -244,7 +248,26 @@ def checkout_project():
return jsonify({'error': 'Failed to clone repository'}), 500 return jsonify({'error': 'Failed to clone repository'}), 500
return jsonify({'error': f"Uncaught error processing checkout: {e}"}), 500 return jsonify({'error': f"Uncaught error processing checkout: {e}"}), 500
finally: finally:
build_lock.release() if not async_mode:
build_lock.release()
def _run_build_background(job_id, project_path, start_time):
"""Run install_and_build in a background thread. Assumes build_lock is held;
the thread releases it when the build finishes."""
def runner():
try:
install_and_build(project_path, job_id, start_time)
except Exception as e:
print(f"Background build failed for {job_id}: {e}")
_clear_progress(job_id)
finally:
build_lock.release()
threading.Thread(target=runner, name=f"build-{job_id}", daemon=True).start()
def _is_async_request():
return request.args.get("async") in ("1", "true", "yes", "on")
def safe_extract(zip_ref, dest_dir): def safe_extract(zip_ref, dest_dir):
"""Extract a zipfile, rejecting members that would escape dest_dir.""" """Extract a zipfile, rejecting members that would escape dest_dir."""
@@ -260,6 +283,8 @@ def upload_project():
if not build_lock.acquire(blocking=False): if not build_lock.acquire(blocking=False):
return jsonify({'error': f"Another build is already running: {system_status['running_job']}"}), 409 return jsonify({'error': f"Another build is already running: {system_status['running_job']}"}), 409
async_mode = _is_async_request()
working_dir = None
try: try:
start_time = datetime.datetime.now() start_time = datetime.datetime.now()
if 'file' not in request.files: if 'file' not in request.files:
@@ -283,15 +308,20 @@ def upload_project():
print(f"Extracting uploaded project zip...") print(f"Extracting uploaded project zip...")
safe_extract(zip_ref, working_dir) safe_extract(zip_ref, working_dir)
if async_mode:
_run_build_background(job_id, working_dir, start_time)
return jsonify({"id": job_id, "async": True, "message": "Build started"}), 202
return install_and_build(working_dir, job_id, start_time) return install_and_build(working_dir, job_id, start_time)
except Exception as e: except Exception as e:
print(f"Uncaught error processing job: {e}") print(f"Uncaught error processing job: {e}")
system_status['status'] = "ready" system_status['status'] = "ready"
_clear_progress(job_id) _clear_progress(job_id)
shutil.rmtree(working_dir, ignore_errors=True) shutil.rmtree(working_dir, ignore_errors=True) if working_dir else None
return jsonify({"error": f"Uncaught error processing job: {e}"}), 500 return jsonify({"error": f"Uncaught error processing job: {e}"}), 500
finally: finally:
build_lock.release() if not async_mode:
build_lock.release()
def install_and_build(project_path, job_id, start_time): def install_and_build(project_path, job_id, start_time):
_set_step(job_id, "starting", project_path=project_path) _set_step(job_id, "starting", project_path=project_path)
+55 -16
View File
@@ -1,6 +1,7 @@
import datetime import datetime
import os import os
import threading import threading
import time
import requests import requests
@@ -97,21 +98,22 @@ class Scheduler:
try: try:
if job["source_type"] == "git": if job["source_type"] == "git":
worker_job = self._request_git(job, worker) worker_job_id = self._request_git(job, worker)
else: else:
worker_job = self._request_upload(job, worker) worker_job_id = self._request_upload(job, worker)
except Exception as e: except Exception as e:
db.update_job(job_id, status="failed", finished_at=datetime.datetime.now().isoformat(), db.update_job(job_id, status="failed", finished_at=datetime.datetime.now().isoformat(),
error=str(e)) error=str(e))
self._append(job_id, f"[scheduler] FAILED: {e}\n") self._append(job_id, f"[scheduler] FAILED: {e}\n")
return return
worker_job_id = worker_job.get("id")
worker_url = worker["url"] worker_url = worker["url"]
db.update_job(job_id, status="building", worker_job_id=worker_job_id, worker_url=worker_url) db.update_job(job_id, status="building", worker_job_id=worker_job_id, worker_url=worker_url)
self._append(job_id, f"[scheduler] Building on {worker_url} (worker job {worker_job_id})\n")
# download artifacts # Poll progress until the worker reports the build finished.
try: try:
self._wait_for_build(job, worker, worker_job_id)
self._fetch_artifacts(job, worker, worker_job_id) self._fetch_artifacts(job, worker, worker_job_id)
db.update_job(job_id, status="done", finished_at=datetime.datetime.now().isoformat()) db.update_job(job_id, status="done", finished_at=datetime.datetime.now().isoformat())
self._append(job_id, "[scheduler] Build complete, artifacts saved.\n") self._append(job_id, "[scheduler] Build complete, artifacts saved.\n")
@@ -123,28 +125,65 @@ class Scheduler:
self._publish(job["id"], "job", '"done"') self._publish(job["id"], "job", '"done"')
def _request_git(self, job, worker): def _request_git(self, job, worker):
self._append(job["id"], f"[scheduler] Cloning {job['source']} on {worker['url']}\n") job_id = job["id"]
resp = requests.post(f"{worker['url']}/checkout_git", self._append(job_id, f"[scheduler] Cloning {job['source']} on {worker['url']}\n")
resp = requests.post(f"{worker['url']}/checkout_git?async=1",
json={"repo_url": job["source"]}, json={"repo_url": job["source"]},
timeout=settings.WORKER_BUILD_TIMEOUT, timeout=settings.WORKER_STATUS_TIMEOUT,
stream=False) stream=False)
return self._worker_response(resp, "checkout") data = self._worker_response(resp, "checkout")
if not data.get("id"):
raise RuntimeError("checkout did not return a job id")
return data["id"]
def _request_upload(self, job, worker): def _request_upload(self, job, worker):
job_id = job["id"]
path = self._source_path(job) path = self._source_path(job)
if not path or not os.path.exists(path): if not path or not os.path.exists(path):
raise FileNotFoundError(f"Source file not found: {path}") raise FileNotFoundError(f"Source file not found: {path}")
self._append(job["id"], f"[scheduler] Uploading {os.path.basename(path)} to {worker['url']}\n") self._append(job_id, f"[scheduler] Uploading {os.path.basename(path)} to {worker['url']}\n")
with open(path, "rb") as f: with open(path, "rb") as f:
resp = requests.post(f"{worker['url']}/upload", resp = requests.post(f"{worker['url']}/upload?async=1",
files={"file": (os.path.basename(path), f, "application/zip")}, files={"file": (os.path.basename(path), f, "application/zip")},
timeout=settings.WORKER_BUILD_TIMEOUT, timeout=settings.WORKER_STATUS_TIMEOUT,
stream=False) stream=False)
worker_job = self._worker_response(resp, "upload") data = self._worker_response(resp, "upload")
if not worker_job.get("id"): if not data.get("id"):
# Agent may have started building before returning; try progress fallback. raise RuntimeError("upload did not return a job id")
pass return data["id"]
return worker_job
def _wait_for_build(self, job, worker, worker_job_id):
"""Poll worker progress until the build completes (progress cleared + worker ready)."""
job_id = job["id"]
deadline = datetime.datetime.now() + datetime.timedelta(seconds=settings.WORKER_BUILD_TIMEOUT)
while datetime.datetime.now() < deadline:
progress = self._probe_progress(job, worker["url"], worker_job_id)
if progress is None:
status = wrk.probe_worker(worker["url"])
if status and status.get("status") == "ready":
return
# progress gone but worker not ready: brief window, keep waiting
time.sleep(settings.SCHEDULER_INTERVAL)
continue
line = progress.get("last_log_line") or progress.get("step") or ""
if line:
self._append(job_id, f"[worker] {line}\n")
time.sleep(settings.SCHEDULER_INTERVAL)
raise RuntimeError(f"Worker did not finish in {settings.WORKER_BUILD_TIMEOUT}s")
@staticmethod
def _probe_progress(job, worker_url, worker_job_id):
try:
resp = requests.get(f"{worker_url}/progress/{worker_job_id}",
timeout=settings.WORKER_STATUS_TIMEOUT)
except requests.RequestException:
return None
if resp.status_code == 404:
return None
try:
return resp.json()
except ValueError:
return None
@staticmethod @staticmethod
def _worker_response(resp, action): def _worker_response(resp, action):