test nightly-distroless

.github/workflows/image.yml

@@ -4,6 +4,7 @@ on:
     branches:
       - master
       - release/**
+      - test/nightly-distroless
 
 concurrency:
   group: ${{ github.workflow }}-${{ github.head_ref || github.run_id }}
@@ -85,6 +86,7 @@ jobs:
         ghcr: true
         tag_nightly: false
         tag_latest: false
+        publish_on_pr: ${{ github.head_ref == 'test/nightly-distroless' }}
 
   # Debug distroless image — with busybox for troubleshooting
   build-debug-multiplatform:
@@ -140,7 +142,7 @@ jobs:
 
   assemble-distroless:
     needs: [build-distroless-multiplatform]
-    if: ${{ (github.ref_name == 'master' || startsWith(github.ref_name, 'release/')) && github.event_name != 'pull_request' }}
+    if: ${{ needs.build-distroless-multiplatform.result == 'success' && github.repository_owner == 'getsentry' && (github.ref_name == 'master' || startsWith(github.ref_name, 'release/') || github.ref_name == 'test/nightly-distroless' || github.head_ref == 'test/nightly-distroless') }}
     runs-on: ubuntu-latest
     permissions:
       contents: read
@@ -155,12 +157,14 @@ jobs:
         uses: docker/setup-buildx-action@8d2750c68a42422c14e847fe6c8ac0403b4cbd6f # v3
 
       - name: Create distroless multiplatform manifests
+        env:
+          IMAGE_SHA: ${{ github.event.pull_request.head.sha || github.sha }}
         run: |
           docker buildx imagetools create \
-            --tag ghcr.io/getsentry/snuba:${{ github.sha }}-distroless \
+            --tag ghcr.io/getsentry/snuba:${IMAGE_SHA}-distroless \
             --tag ghcr.io/getsentry/snuba:nightly-distroless \
-            ghcr.io/getsentry/snuba:${{ github.sha }}-distroless-amd64 \
-            ghcr.io/getsentry/snuba:${{ github.sha }}-distroless-arm64
+            ghcr.io/getsentry/snuba:${IMAGE_SHA}-distroless-amd64 \
+            ghcr.io/getsentry/snuba:${IMAGE_SHA}-distroless-arm64
 
   assemble-debug:
     needs: [build-debug-multiplatform]

devservices/config.yml

@@ -75,15 +75,15 @@ services:
     restart: unless-stopped
 
   snuba:
-    image: ghcr.io/getsentry/snuba:nightly
+    image: ghcr.io/getsentry/snuba:nightly-distroless
     ports:
       - 127.0.0.1:1218:1218
       - 127.0.0.1:1219:1219
     command:
       - devserver
       - --${SNUBA_NO_WORKERS:+no-workers}
     healthcheck:
-      test: curl -f http://localhost:1218/health_envoy
+      test: ["CMD", "python3", "-c", "import urllib.request; urllib.request.urlopen('http://localhost:1218/health_envoy')"]
       interval: 5s
       timeout: 5s
       retries: 3
@@ -114,7 +114,7 @@ services:
       - orchestrator=devservices
     restart: unless-stopped
   profiles-consumer:
-    image: ghcr.io/getsentry/snuba:nightly
+    image: ghcr.io/getsentry/snuba:nightly-distroless
     command: [
           rust-consumer,
           --storage=profiles,
@@ -143,7 +143,7 @@ services:
       - orchestrator=devservices
     restart: unless-stopped
   profile-chunks-consumer:
-    image: ghcr.io/getsentry/snuba:nightly
+    image: ghcr.io/getsentry/snuba:nightly-distroless
     command: [
           rust-consumer,
           --storage=profile_chunks,
@@ -172,7 +172,7 @@ services:
       - orchestrator=devservices
     restart: unless-stopped
   functions-consumer:
-    image: ghcr.io/getsentry/snuba:nightly
+    image: ghcr.io/getsentry/snuba:nightly-distroless
     command: [
           rust-consumer,
           --storage=functions_raw,
@@ -201,7 +201,7 @@ services:
       - orchestrator=devservices
     restart: unless-stopped
   metrics-consumer:
-    image: ghcr.io/getsentry/snuba:nightly
+    image: ghcr.io/getsentry/snuba:nightly-distroless
     command: [
           rust-consumer,
           --storage=metrics_raw,
@@ -230,7 +230,7 @@ services:
       - orchestrator=devservices
     restart: unless-stopped
   generic-metrics-distributions-consumer:
-    image: ghcr.io/getsentry/snuba:nightly
+    image: ghcr.io/getsentry/snuba:nightly-distroless
     command: [
           rust-consumer,
           --storage=generic_metrics_distributions_raw,
@@ -259,7 +259,7 @@ services:
       - orchestrator=devservices
     restart: unless-stopped
   generic-metrics-sets-consumer:
-    image: ghcr.io/getsentry/snuba:nightly
+    image: ghcr.io/getsentry/snuba:nightly-distroless
     command: [
           rust-consumer,
           --storage=generic_metrics_sets_raw,
@@ -288,7 +288,7 @@ services:
       - orchestrator=devservices
     restart: unless-stopped
   generic-metrics-counters-consumer:
-    image: ghcr.io/getsentry/snuba:nightly
+    image: ghcr.io/getsentry/snuba:nightly-distroless
     command: [
           rust-consumer,
           --storage=generic_metrics_counters_raw,
@@ -317,7 +317,7 @@ services:
       - orchestrator=devservices
     restart: unless-stopped
   generic-metrics-gauges-consumer:
-    image: ghcr.io/getsentry/snuba:nightly
+    image: ghcr.io/getsentry/snuba:nightly-distroless
     command: [
           rust-consumer,
           --storage=generic_metrics_gauges_raw,

granian_entrypoint.py

@@ -0,0 +1,42 @@
+#!/usr/bin/env python3
+"""
+Python replacement for granian_entrypoint.sh — compatible with distroless images.
+"""
+
+import os
+import time
+import urllib.error
+import urllib.request
+
+
+def wait_for_envoy() -> None:
+    port = os.environ.get("ENVOY_ADMIN_PORT")
+    if not port:
+        print("ENVOY_ADMIN_PORT env var is unset")
+        print("Fall through to snuba start without check")
+        return
+
+    url = f"http://localhost:{port}/ready"
+    print(f"Check envoy readiness on {url}")
+    while True:
+        time.sleep(1)
+        try:
+            with urllib.request.urlopen(url, timeout=2) as resp:
+                if resp.read().decode().strip() == "LIVE":
+                    print("Envoy is ready")
+                    return
+        except Exception:
+            pass
+        print("Envoy not ready, looping..")
+
+
+def main() -> None:
+    wait_for_envoy()
+
+    wsgi_target = os.environ.get("SNUBA_WSGI_TARGET", "snuba.web.wsgi:application")
+    args = ["granian", "--interface", "wsgi", "--host", "0.0.0.0", "--http", "auto", wsgi_target]
+    os.execvp(args[0], args)
+
+
+if __name__ == "__main__":
+    main()

pyproject.toml

@@ -78,7 +78,6 @@ snuba = "snuba.cli:main"
 dev = [
     "devservices>=1.2.1",
     "freezegun>=1.5.5",
-    "honcho>=1.1.0",
     "mypy>=1.1.1",
     "pre-commit>=4.2.0",
     "pytest>=8.3.3",
@@ -136,7 +135,6 @@ module = [
     "fastjsonschema",
     "fastjsonschema.exceptions",
     "granian",
-    "honcho.manager",
     "jsonschema",
     "jsonschema.exceptions",
     "jsonschema2md",

snuba/cli/devserver.py

@@ -1,11 +1,29 @@
 import os
+import signal
+import subprocess
 import sys
-from subprocess import call, list2cmdline
+import threading
+from subprocess import call
 
 import click
 
 from snuba import settings
 
+# Match honcho: SIGTERM, then SIGKILL if children do not exit (avoids indefinite
+# hang when a child ignores SIGTERM; see also PEP 475 / wait() retry after signals).
+_SUBPROCESS_TERM_GRACE_SEC = 5.0
+
+
+def _reap_after_terminate(proc: subprocess.Popen[bytes], grace_sec: float) -> None:
+    """Wait for proc to exit after terminate(); kill -9 if still alive after grace_sec."""
+    try:
+        proc.wait(timeout=grace_sec)
+    except subprocess.TimeoutExpired:
+        if proc.poll() is None:
+            proc.kill()
+            proc.wait()
+
+
 COMMON_RUST_CONSUMER_DEV_OPTIONS = [
     "--use-rust-processor",
     "--auto-offset-reset=latest",
@@ -21,8 +39,6 @@
 def devserver(*, bootstrap: bool, workers: bool, log_level: str) -> None:
     "Starts all Snuba processes for local development."
 
-    from honcho.manager import Manager
-
     os.environ["PYTHONUNBUFFERED"] = "1"
 
     if bootstrap:
@@ -518,13 +534,82 @@ def devserver(*, bootstrap: bool, workers: bool, log_level: str) -> None:
             ),
         ]
 
-    manager = Manager()
+    sys.exit(_run_daemons(daemons))
+
+
+def _run_daemons(daemons: list[tuple[str, list[str]]]) -> int:
+    procs: dict[str, subprocess.Popen[bytes]] = {}
+    threads: list[threading.Thread] = []
+    first_failure: list[int] = []
+    done = threading.Event()
+    cleanup_started = threading.Event()
+    failure_lock = threading.Lock()
+    supervisor_signal: list[int] = []
+
+    def shutdown(signum: int, frame: object) -> None:
+        # Mark cleanup before terminate so stream threads do not treat SIGTERM as a
+        # natural crash (honcho parity when one daemon exits or user interrupts).
+        cleanup_started.set()
+        if not supervisor_signal:
+            supervisor_signal.append(signum)
+        for proc in procs.values():
+            if proc.poll() is None:
+                proc.terminate()
+        done.set()
+
+    signal.signal(signal.SIGINT, shutdown)
+    signal.signal(signal.SIGTERM, shutdown)
+
+    def stream(name: str, proc: subprocess.Popen[bytes]) -> None:
+        try:
+            assert proc.stdout is not None
+            for line in proc.stdout:
+                sys.stdout.write(f"{name} | {line.decode(errors='replace')}")
+                sys.stdout.flush()
+            rc = proc.wait()
+            with failure_lock:
+                if rc != 0 and not cleanup_started.is_set():
+                    if not first_failure:
+                        first_failure.append(rc)
+        except BaseException:
+            with failure_lock:
+                if not cleanup_started.is_set() and not first_failure:
+                    first_failure.append(1)
+            raise
+        finally:
+            # Always unblock the supervisor (e.g. BrokenPipe/EPIPE on stdout write).
+            done.set()
+
     for name, cmd in daemons:
-        manager.add_process(
-            name,
-            list2cmdline(cmd),
-            quiet=False,
+        proc = subprocess.Popen(
+            cmd,
+            stdin=subprocess.DEVNULL,
+            stdout=subprocess.PIPE,
+            stderr=subprocess.STDOUT,
         )
+        procs[name] = proc
+        t = threading.Thread(target=stream, args=(name, proc), daemon=True)
+        t.start()
+        threads.append(t)
+
+    done.wait()
+    cleanup_started.set()
+    # Any daemon exit ends the supervisor; terminate the rest (honcho parity).
+    for proc in procs.values():
+        if proc.poll() is None:
+            proc.terminate()
+
+    for proc in procs.values():
+        if proc.poll() is None:
+            _reap_after_terminate(proc, _SUBPROCESS_TERM_GRACE_SEC)
+        else:
+            proc.wait()
+
+    for t in threads:
+        t.join()
 
-    manager.loop()
-    sys.exit(manager.returncode)
+    if first_failure:
+        return first_failure[0]
+    if supervisor_signal:
+        return 128 + supervisor_signal[0]
+    return 0