gowalk-cicd 1.0.196 → 1.0.198
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/backend-action/action.yml +6 -2
- package/backend-action/identity_diagnostics.py +17 -9
- package/backend-action/runtime_diagnostics.py +35 -4
- package/backend-action/runtime_image_health.py +10 -1
- package/docs/BACKEND_HEALTHCHECKS.md +4 -0
- package/docs/BACKEND_SSH_GATE.md +13 -1
- package/package.json +1 -1
|
@@ -306,9 +306,11 @@ runs:
|
|
|
306
306
|
run: |
|
|
307
307
|
RESULT_DIR="$(mktemp -d "$RUNNER_TEMP/backend-runtime-diagnostics.XXXXXX")"
|
|
308
308
|
echo "result=$RESULT_DIR/result.json" >> "$GITHUB_OUTPUT"
|
|
309
|
+
echo "transport=$RESULT_DIR/transport.json" >> "$GITHUB_OUTPUT"
|
|
309
310
|
python3 "${{ github.action_path }}/runtime_diagnostics.py" \
|
|
310
311
|
--app "$DIAGNOSTIC_APP" --service "$DIAGNOSTIC_SERVICE" --expected-build-sha "$DIAGNOSTIC_BUILD" \
|
|
311
|
-
--host "$DIAGNOSTIC_HOST" --user "$DIAGNOSTIC_USER" --output "$RESULT_DIR/result.json"
|
|
312
|
+
--host "$DIAGNOSTIC_HOST" --user "$DIAGNOSTIC_USER" --output "$RESULT_DIR/result.json" \
|
|
313
|
+
--transport-output "$RESULT_DIR/transport.json"
|
|
312
314
|
|
|
313
315
|
- name: Retain scoped backend runtime result
|
|
314
316
|
if: always() && inputs.operation == 'runtime-diagnostics'
|
|
@@ -317,7 +319,9 @@ runs:
|
|
|
317
319
|
NODE_OPTIONS: --require "${{ github.action_path }}/scripts/artifact_env.cjs"
|
|
318
320
|
with:
|
|
319
321
|
name: backend-runtime-diagnostics-${{ github.job }}-${{ github.run_attempt }}
|
|
320
|
-
path:
|
|
322
|
+
path: |
|
|
323
|
+
${{ steps.runtime_diagnostics.outputs.result }}
|
|
324
|
+
${{ steps.runtime_diagnostics.outputs.transport }}
|
|
321
325
|
if-no-files-found: error
|
|
322
326
|
|
|
323
327
|
- name: Read scoped backend network prerequisites
|
|
@@ -40,7 +40,7 @@ class ReaderFailure(Exception):
|
|
|
40
40
|
super().__init__(self.code)
|
|
41
41
|
|
|
42
42
|
|
|
43
|
-
def transfer(selector, process, payload, offset, data):
|
|
43
|
+
def transfer(selector, process, payload, offset, data, errors):
|
|
44
44
|
for key, _ in selector.select(0.1):
|
|
45
45
|
if key.fileobj is process.stdin:
|
|
46
46
|
offset += os.write(process.stdin.fileno(), payload[offset:offset + 4096])
|
|
@@ -48,23 +48,27 @@ def transfer(selector, process, payload, offset, data):
|
|
|
48
48
|
selector.unregister(process.stdin)
|
|
49
49
|
process.stdin.close()
|
|
50
50
|
else:
|
|
51
|
-
chunk = os.read(
|
|
51
|
+
chunk = os.read(key.fileobj.fileno(), 65536)
|
|
52
52
|
if not chunk:
|
|
53
|
-
selector.unregister(
|
|
54
|
-
data.extend(chunk)
|
|
55
|
-
if len(data) > LIMIT:
|
|
53
|
+
selector.unregister(key.fileobj)
|
|
54
|
+
(errors if key.fileobj is process.stderr else data).extend(chunk)
|
|
55
|
+
if len(data) + len(errors) > LIMIT:
|
|
56
56
|
raise ReaderFailure('output_limit')
|
|
57
57
|
return offset
|
|
58
58
|
|
|
59
59
|
|
|
60
|
-
def bounded(argv, *, payload=None, seconds=30, with_status=False):
|
|
60
|
+
def bounded(argv, *, payload=None, seconds=30, with_status=False, with_stderr=False):
|
|
61
61
|
"""Bound output, runtime and owned process cleanup without printing stderr."""
|
|
62
62
|
process = subprocess.Popen(argv, stdin=subprocess.PIPE if payload else subprocess.DEVNULL,
|
|
63
|
-
stdout=subprocess.PIPE,
|
|
64
|
-
|
|
63
|
+
stdout=subprocess.PIPE,
|
|
64
|
+
stderr=subprocess.PIPE if with_stderr else subprocess.DEVNULL,
|
|
65
|
+
start_new_session=True)
|
|
66
|
+
deadline, data, errors = time.monotonic() + seconds, bytearray(), bytearray()
|
|
65
67
|
try:
|
|
66
68
|
with selectors.DefaultSelector() as selector:
|
|
67
69
|
selector.register(process.stdout, selectors.EVENT_READ)
|
|
70
|
+
if with_stderr:
|
|
71
|
+
selector.register(process.stderr, selectors.EVENT_READ)
|
|
68
72
|
if payload:
|
|
69
73
|
os.set_blocking(process.stdin.fileno(), False)
|
|
70
74
|
selector.register(process.stdin, selectors.EVENT_WRITE)
|
|
@@ -72,12 +76,14 @@ def bounded(argv, *, payload=None, seconds=30, with_status=False):
|
|
|
72
76
|
while selector.get_map():
|
|
73
77
|
if time.monotonic() >= deadline:
|
|
74
78
|
raise ReaderFailure('command_timeout')
|
|
75
|
-
offset = transfer(selector, process, payload, offset, data)
|
|
79
|
+
offset = transfer(selector, process, payload, offset, data, errors)
|
|
76
80
|
try:
|
|
77
81
|
status = process.wait(timeout=max(0.01, deadline - time.monotonic()))
|
|
78
82
|
except subprocess.TimeoutExpired:
|
|
79
83
|
raise ReaderFailure('command_timeout') from None
|
|
80
84
|
if with_status:
|
|
85
|
+
if with_stderr:
|
|
86
|
+
return status, bytes(data), bytes(errors)
|
|
81
87
|
return status, bytes(data)
|
|
82
88
|
if status:
|
|
83
89
|
raise ReaderFailure('command_failed')
|
|
@@ -90,6 +96,8 @@ def bounded(argv, *, payload=None, seconds=30, with_status=False):
|
|
|
90
96
|
pass
|
|
91
97
|
process.wait(timeout=5)
|
|
92
98
|
process.stdout.close()
|
|
99
|
+
if process.stderr:
|
|
100
|
+
process.stderr.close()
|
|
93
101
|
if process.stdin and not process.stdin.closed:
|
|
94
102
|
process.stdin.close()
|
|
95
103
|
|
|
@@ -3,27 +3,50 @@ import argparse
|
|
|
3
3
|
import json
|
|
4
4
|
import re
|
|
5
5
|
import shlex
|
|
6
|
+
import sys
|
|
6
7
|
from pathlib import Path
|
|
7
8
|
|
|
8
9
|
from identity_diagnostics import ReaderFailure, bounded, retain_result
|
|
9
10
|
from runtime_projection import RuntimeFailure, envelope, scope, validate
|
|
10
11
|
|
|
11
12
|
|
|
12
|
-
def
|
|
13
|
+
def transport_signals(status, raw):
|
|
14
|
+
"""Observe fixed SSH failure labels; private stderr never enters a receipt."""
|
|
15
|
+
lines = raw.decode('utf-8', errors='replace').splitlines()
|
|
16
|
+
patterns = {
|
|
17
|
+
1: [('gate_refused', r'backend_ssh_refused')],
|
|
18
|
+
127: [('remote_command_unavailable',
|
|
19
|
+
r'(?:/bin/)?(?:bash|sh): (?:(?:line )?\d+: )?gowalk-backend: (?:command )?not found')],
|
|
20
|
+
255: [
|
|
21
|
+
('authentication_refused', r'[^\r\n]+: Permission denied \([a-z,-]+\)\.'),
|
|
22
|
+
('host_key_refused', r'Host key verification failed\.'),
|
|
23
|
+
('connection_refused', r'ssh: connect to host [^\r\n ]+ port \d+: Connection refused'),
|
|
24
|
+
('connection_timeout', r'ssh: connect to host [^\r\n ]+ port \d+: (?:Connection|Operation) timed out'),
|
|
25
|
+
('name_resolution_failed', r'ssh: Could not resolve hostname [^\r\n ]+: [^\r\n]+'),
|
|
26
|
+
],
|
|
27
|
+
}
|
|
28
|
+
return [name for name, pattern in patterns.get(status, [])
|
|
29
|
+
if any(re.fullmatch(pattern, line) for line in lines)]
|
|
30
|
+
|
|
31
|
+
|
|
32
|
+
def local(args, transport=None):
|
|
13
33
|
scope(args.app, args.service, args.expected_build_sha)
|
|
14
34
|
if (not re.fullmatch(r'[A-Za-z0-9][A-Za-z0-9.-]{0,252}', args.host or '')
|
|
15
35
|
or not re.fullmatch(r'[a-z_][a-z0-9_-]{0,31}', args.user)):
|
|
16
36
|
raise RuntimeFailure('invalid_scope')
|
|
17
37
|
command = shlex.join(['gowalk-backend', 'runtime-diagnostics', args.app, args.service, args.expected_build_sha])
|
|
18
38
|
try:
|
|
19
|
-
status, raw = bounded(['ssh', '-T', '-i', str(Path.home() / '.ssh/backend_deploy'),
|
|
39
|
+
status, raw, errors = bounded(['ssh', '-T', '-i', str(Path.home() / '.ssh/backend_deploy'),
|
|
20
40
|
'-o', 'IdentitiesOnly=yes', '-o', 'BatchMode=yes', '-o', 'StrictHostKeyChecking=yes',
|
|
21
41
|
'-o', 'ConnectTimeout=10', '-o', 'ServerAliveInterval=5', '-o', 'ServerAliveCountMax=2',
|
|
22
|
-
'-o', 'ClearAllForwardings=yes', args.user + '@' + args.host, command],
|
|
42
|
+
'-o', 'ClearAllForwardings=yes', args.user + '@' + args.host, command],
|
|
43
|
+
seconds=75, with_status=True, with_stderr=True)
|
|
23
44
|
except ReaderFailure as error:
|
|
24
45
|
raise RuntimeFailure(error.code) from None
|
|
25
46
|
except OSError:
|
|
26
47
|
raise RuntimeFailure('transport_failure') from None
|
|
48
|
+
if transport is not None:
|
|
49
|
+
transport.update(exit_status=status, signals=transport_signals(status, errors))
|
|
27
50
|
if status not in (0, 1):
|
|
28
51
|
raise RuntimeFailure('transport_failure')
|
|
29
52
|
if status == 1 and not raw.strip():
|
|
@@ -45,14 +68,22 @@ def main():
|
|
|
45
68
|
parser.add_argument('--host', required=True)
|
|
46
69
|
parser.add_argument('--user', default='root')
|
|
47
70
|
parser.add_argument('--output', required=True)
|
|
71
|
+
parser.add_argument('--transport-output', help='Separate closed SSH observation receipt')
|
|
48
72
|
args = parser.parse_args()
|
|
73
|
+
transport = {'schema': 'gowalk-cicd/backend-runtime-transport.v1', 'exit_status': None, 'signals': []}
|
|
49
74
|
try:
|
|
50
|
-
result = local(args)
|
|
75
|
+
result = local(args, transport)
|
|
51
76
|
except Exception as error:
|
|
52
77
|
result = envelope(args.app, args.service, args.expected_build_sha,
|
|
53
78
|
error.code if isinstance(error, RuntimeFailure) else 'unclassified', reader_version=None)
|
|
54
79
|
try:
|
|
55
80
|
retain_result(args.output, result)
|
|
81
|
+
if args.transport_output:
|
|
82
|
+
retain_result(args.transport_output, transport)
|
|
83
|
+
status = 'unknown' if transport['exit_status'] is None else str(transport['exit_status'])
|
|
84
|
+
print('::notice title=backend_runtime_transport::schema=' + transport['schema']
|
|
85
|
+
+ ' exit_status=' + status + ' signals=' + (','.join(transport['signals']) or 'none'),
|
|
86
|
+
file=sys.stderr)
|
|
56
87
|
except ReaderFailure:
|
|
57
88
|
result = envelope(args.app, args.service, args.expected_build_sha, 'artifact_write_failed', reader_version=None)
|
|
58
89
|
print(json.dumps(result, separators=(',', ':')))
|
|
@@ -5,6 +5,7 @@ import re
|
|
|
5
5
|
import subprocess
|
|
6
6
|
|
|
7
7
|
from runtime_healthcheck import bounded_tmpfs, cache_healthcheck
|
|
8
|
+
from runtime_model import escape_interpolation
|
|
8
9
|
|
|
9
10
|
|
|
10
11
|
def eligible(service):
|
|
@@ -17,7 +18,14 @@ def eligible(service):
|
|
|
17
18
|
def image_health(command, name, service):
|
|
18
19
|
reference = service.get('image')
|
|
19
20
|
if not reference:
|
|
20
|
-
|
|
21
|
+
resolver = list(command)
|
|
22
|
+
option = next((flag for flag in ('--file', '-f') if flag in resolver), None)
|
|
23
|
+
if option:
|
|
24
|
+
resolver[resolver.index(option) + 1] = '-'
|
|
25
|
+
else:
|
|
26
|
+
resolver.extend(['--file', '-'])
|
|
27
|
+
selected = {'services': {name: {'build': '.'}}}
|
|
28
|
+
result = subprocess.run(resolver + ['config', '--images', name], input=json.dumps(selected),
|
|
21
29
|
capture_output=True, text=True, timeout=30, check=True)
|
|
22
30
|
references = result.stdout.split()
|
|
23
31
|
if len(references) != 1:
|
|
@@ -44,6 +52,7 @@ def inherit_healthchecks(model, command):
|
|
|
44
52
|
candidate = copy.deepcopy(service)
|
|
45
53
|
candidate.setdefault('healthcheck', {})['test'] = health['Test']
|
|
46
54
|
if cache_healthcheck(candidate):
|
|
55
|
+
candidate['healthcheck']['test'] = escape_interpolation(candidate['healthcheck']['test'])
|
|
47
56
|
service['healthcheck'] = candidate['healthcheck']
|
|
48
57
|
service.update(image=image['Id'], pull_policy='never')
|
|
49
58
|
changed = True
|
|
@@ -36,6 +36,10 @@ wraps only a missing Compose test, retaining explicit timing and disabled/custom
|
|
|
36
36
|
launch model pins the transformed service to that inspected image ID with pulling disabled, so a
|
|
37
37
|
moving tag cannot substitute another image's healthcheck. Explicit registry-refresh policies and
|
|
38
38
|
non-build services retain their existing image/pull behavior. No pre-build image healthcheck is reused.
|
|
39
|
+
Imported image commands escape literal dollars once at the Compose boundary, preserving Python
|
|
40
|
+
strings and container-shell expansion without interpreting them against the deployment host environment.
|
|
41
|
+
Default image names resolve from an isolated service model with the same Compose project name;
|
|
42
|
+
dependency images cannot make that selection ambiguous.
|
|
39
43
|
|
|
40
44
|
## Adopt an existing API
|
|
41
45
|
|
package/docs/BACKEND_SSH_GATE.md
CHANGED
|
@@ -68,7 +68,7 @@ consumer code or an uploaded script. An older/refusing gate has no reader receip
|
|
|
68
68
|
`runtime_reader_unavailable`; locally synthesized failures carry `reader_version: null`, never host-adoption proof.
|
|
69
69
|
A collector refusal carries its actual `reader_version` and fixed failure code.
|
|
70
70
|
|
|
71
|
-
The `backend-runtime-diagnostics-<job>-<attempt>` artifact contains
|
|
71
|
+
The `backend-runtime-diagnostics-<job>-<attempt>` artifact contains `result.json`, schema
|
|
72
72
|
`gowalk-cicd/backend-runtime-diagnostics.v1`. Success binds the expected/observed build to the running API's
|
|
73
73
|
`BUILD_SHA`, enumerates at most 32 exact-project containers (including stopped ones), and rechecks their exact
|
|
74
74
|
project/service/working-directory labels. It returns container and image IDs, effective memory/reservation/CPU/PID
|
|
@@ -77,6 +77,18 @@ Docker's display precision. CPU samples are Docker's bounded observations, not a
|
|
|
77
77
|
collector never launches workload, reads logs, executes in containers or inspects a shared database's cgroup.
|
|
78
78
|
Both container identity/configuration and the private profile are rechecked before returning success.
|
|
79
79
|
|
|
80
|
+
The same artifact also retains `transport.json`, schema `gowalk-cicd/backend-runtime-transport.v1`.
|
|
81
|
+
Its fixed fields also appear in the brace-free `backend_runtime_transport` notice. It reports the SSH child's
|
|
82
|
+
`exit_status` (null when no complete exit was observed) and fixed `signals`: `gate_refused`,
|
|
83
|
+
`remote_command_unavailable`, `authentication_refused`, `host_key_refused`, `connection_refused`,
|
|
84
|
+
`connection_timeout` or `name_resolution_failed`. Unknown text yields no signal; stderr and remote
|
|
85
|
+
command output never enter this receipt. The original `result.json` schema and failure codes remain
|
|
86
|
+
unchanged. These observations explain a refused read, never attest host reader installation or authorize
|
|
87
|
+
a retry or transport bypass. Exit 127 without the exact command-not-found observation does not prove a
|
|
88
|
+
missing reader; a gate refusal does not identify its policy or ownership cause. Normal backend deployment
|
|
89
|
+
does not install the trusted reader, and an updated workflow client alone cannot repair an unavailable
|
|
90
|
+
host command. Install the published host runtime through the procedure below.
|
|
91
|
+
|
|
80
92
|
`database.mode: host_profile` with `binding_matches_profile: true` proves this app's root-owned profile supplied
|
|
81
93
|
the API's current private `DATABASE_URL`. `removed_services` describes that profile's applied removals; a remaining
|
|
82
94
|
removed service refuses. `app_service` means the runtime connection names another reported service.
|