fix(ops): bound CD DB probes and restore SigNoz collector

This commit is contained in:
ogt
2026-07-15 02:29:51 +08:00
parent 27736add14
commit bfb2d2d040
11 changed files with 442 additions and 19 deletions

View File

@@ -1,10 +1,11 @@
#!/bin/bash
# =============================================================================
# WOOO AIOps - SignOz 備份腳本 (ClickHouse + SQLite)
# 版本: 1.1.0
# 版本: 1.2.0
# 建立日期: 2026-04-05
# 2026-04-05 Claude Code: 新增 SignOz 分散式追蹤備份 — 首席架構師備份審計
# 2026-04-05 Claude Code: v1.1 修正 tar pipeline exit code 處理 + || true
# 2026-07-15 Codex: v1.2 確保任何 exit/signal/failure 都會受控還原 OTEL Collector
# =============================================================================
set -euo pipefail
@@ -14,11 +15,42 @@ source "$(dirname "$0")/common.sh"
SERVICE="signoz"
LOCAL_REPO="${BACKUP_BASE}/signoz"
DUMP_DIR="/tmp/signoz-backup-$$"
COLLECTOR_RESTORE_PENDING=0
CLEANUP_COMPLETE=0
restore_collector() {
if [ "${COLLECTOR_RESTORE_PENDING}" -ne 1 ]; then
return 0
fi
log_info "重啟 signoz-otel-collector..."
if docker start signoz-otel-collector 2>/dev/null; then
COLLECTOR_RESTORE_PENDING=0
return 0
fi
log_warn "signoz-otel-collector 重啟失敗"
return 1
}
cleanup() {
# 確保 collector 已重啟
docker start signoz-otel-collector 2>/dev/null || true
local exit_code=$?
if [ "${CLEANUP_COMPLETE}" -eq 1 ]; then
return "${exit_code}"
fi
CLEANUP_COMPLETE=1
# cleanup 期間忽略重複 signal避免還原與刪除流程重入。
trap '' HUP INT TERM
restore_collector || true
rm -rf "${DUMP_DIR}"
return "${exit_code}"
}
on_signal() {
local signal_number="$1"
exit "$((128 + signal_number))"
}
backup_volume() {
@@ -44,6 +76,12 @@ backup_volume() {
main() {
local start_time=$(date +%s)
trap cleanup EXIT
trap 'on_signal 1' HUP
trap 'on_signal 2' INT
trap 'on_signal 15' TERM
log_info "========== 開始 SignOz 備份 =========="
mkdir -p "${DUMP_DIR}"
@@ -51,13 +89,14 @@ main() {
# Step 1: 停止 OTEL Collector 確保數據一致性
log_info "暫停 signoz-otel-collector 以確保數據一致性..."
# 先 arm 還原再停止,消除 signal 發生在 stop 與狀態記錄之間的空窗。
COLLECTOR_RESTORE_PENDING=1
docker stop signoz-otel-collector 2>/dev/null || log_warn "signoz-otel-collector 未在運行,繼續"
docker stop signoz-telemetrystore-migrator 2>/dev/null || true
# Step 2: 備份 ClickHouse volume (排除 tmp 目錄降低體積)
backup_volume "signoz-clickhouse" "${DUMP_DIR}/clickhouse_${timestamp}.tar.gz" "--exclude=/data/tmp" || {
log_error "ClickHouse volume 備份失敗"
cleanup
notify_clawbot "failed" "${SERVICE}" "SignOz ClickHouse 備份失敗"
exit 1
}
@@ -68,15 +107,14 @@ main() {
}
# Step 4: 重啟 Collector
log_info "重啟 signoz-otel-collector..."
docker start signoz-otel-collector 2>/dev/null || log_warn "signoz-otel-collector 重啟失敗"
# 若此處失敗,保留 pending 讓 EXIT cleanup 再做一次有界還原。
restore_collector || true
# Step 5: 初始化 Restic 倉庫
if [ ! -d "${LOCAL_REPO}/data" ]; then
log_info "初始化 Restic 倉庫: ${LOCAL_REPO}"
restic -r "${LOCAL_REPO}" init --password-file "${RESTIC_PASSWORD_FILE}" 2>&1 || {
log_error "Restic 倉庫初始化失敗"
rm -rf "${DUMP_DIR}"
exit 1
}
fi
@@ -92,8 +130,6 @@ main() {
# Step 7: GFS 清理
cleanup_old_backups "${LOCAL_REPO}"
rm -rf "${DUMP_DIR}"
local end_time=$(date +%s)
local duration=$((end_time - start_time))
log_success "========== SignOz 備份完成 (${duration}s) =========="

View File

@@ -0,0 +1,179 @@
from __future__ import annotations
import os
import shutil
import subprocess
from pathlib import Path
ROOT = Path(__file__).resolve().parents[3]
SCRIPT = ROOT / "scripts" / "backup" / "backup-signoz.sh"
def _write_executable(path: Path, text: str) -> None:
path.write_text(text, encoding="utf-8")
path.chmod(0o755)
def _run_backup(
tmp_path: Path,
*,
docker_run_mode: str = "success",
docker_start_fail_once: bool = False,
restic_backup_status: int = 0,
) -> tuple[subprocess.CompletedProcess[str], list[str], Path]:
harness = tmp_path / "backup"
fake_bin = tmp_path / "bin"
backup_base = tmp_path / "repository"
docker_log = tmp_path / "docker.log"
harness.mkdir()
fake_bin.mkdir()
(backup_base / "signoz" / "data").mkdir(parents=True)
shutil.copy2(SCRIPT, harness / SCRIPT.name)
(harness / "common.sh").write_text(
"""\
BACKUP_BASE="${TEST_BACKUP_BASE:?}"
RESTIC_PASSWORD_FILE="${TEST_BACKUP_BASE}/password"
log_info() { :; }
log_warn() { :; }
log_error() { :; }
log_success() { :; }
notify_clawbot() { :; }
build_tags() { :; }
cleanup_old_backups() { :; }
""",
encoding="utf-8",
)
_write_executable(
fake_bin / "docker",
"""\
#!/bin/bash
set -eu
printf '%s\n' "$*" >> "${TEST_DOCKER_LOG:?}"
case "${1:-}" in
stop)
exit 0
;;
start)
if [ "${FAKE_DOCKER_START_FAIL_ONCE:-0}" = "1" ] && \
[ ! -e "${TEST_DOCKER_START_MARKER:?}" ]; then
: > "${TEST_DOCKER_START_MARKER}"
exit 1
fi
exit 0
;;
run)
if [ "${FAKE_DOCKER_RUN_MODE:-success}" = "signal" ]; then
kill -TERM "${PPID}"
sleep 0.1
exit 0
fi
if [ "${FAKE_DOCKER_RUN_MODE:-success}" = "empty" ]; then
exit 0
fi
printf 'archive-data\n'
;;
esac
""",
)
_write_executable(
fake_bin / "restic",
"""\
#!/bin/bash
set -eu
case " $* " in
*" backup "*) exit "${FAKE_RESTIC_BACKUP_STATUS:-0}" ;;
*" snapshots "*) printf '[{"short_id":"test123"}]\n' ;;
esac
""",
)
_write_executable(
fake_bin / "du",
"""\
#!/bin/bash
printf '4K\t%s\n' "${2:-unknown}"
""",
)
env = {
**os.environ,
"PATH": f"{fake_bin}:{os.environ['PATH']}",
"TEST_BACKUP_BASE": str(backup_base),
"TEST_DOCKER_LOG": str(docker_log),
"TEST_DOCKER_START_MARKER": str(tmp_path / "docker-start-failed-once"),
"FAKE_DOCKER_RUN_MODE": docker_run_mode,
"FAKE_DOCKER_START_FAIL_ONCE": "1" if docker_start_fail_once else "0",
"FAKE_RESTIC_BACKUP_STATUS": str(restic_backup_status),
}
process = subprocess.Popen(
["bash", str(harness / SCRIPT.name)],
env=env,
text=True,
stdout=subprocess.PIPE,
stderr=subprocess.PIPE,
)
stdout, stderr = process.communicate(timeout=10)
result = subprocess.CompletedProcess(
process.args,
process.returncode,
stdout,
stderr,
)
docker_calls = docker_log.read_text(encoding="utf-8").splitlines()
return result, docker_calls, Path(f"/tmp/signoz-backup-{process.pid}")
def _collector_start_count(docker_calls: list[str]) -> int:
return docker_calls.count("start signoz-otel-collector")
def test_collector_is_restored_once_when_backup_receives_term(tmp_path: Path) -> None:
result, docker_calls, dump_dir = _run_backup(
tmp_path,
docker_run_mode="signal",
)
assert result.returncode != 0
assert "stop signoz-otel-collector" in docker_calls
assert _collector_start_count(docker_calls) == 1
assert not dump_dir.exists()
def test_collector_is_restored_once_when_clickhouse_backup_fails(tmp_path: Path) -> None:
result, docker_calls, dump_dir = _run_backup(
tmp_path,
docker_run_mode="empty",
)
assert result.returncode == 1
assert _collector_start_count(docker_calls) == 1
assert not dump_dir.exists()
def test_exit_cleanup_does_not_restart_an_already_restored_collector(
tmp_path: Path,
) -> None:
result, docker_calls, dump_dir = _run_backup(
tmp_path,
restic_backup_status=17,
)
assert result.returncode == 17
assert _collector_start_count(docker_calls) == 1
assert not dump_dir.exists()
def test_exit_cleanup_retries_only_after_explicit_restore_fails(
tmp_path: Path,
) -> None:
result, docker_calls, dump_dir = _run_backup(
tmp_path,
docker_start_fail_once=True,
restic_backup_status=17,
)
assert result.returncode == 17
assert _collector_start_count(docker_calls) == 2
assert not dump_dir.exists()

View File

@@ -22,6 +22,9 @@ EXPECTED_BLOCKERS = [
"direct_ingress_policy_not_closed",
"http_bridge_not_supervised",
"filelog_normalization_degraded",
"api_runtime_cooldown_pending",
"cd_probe_safety_fix_not_deployed",
"backup_collector_restore_guard_not_deployed",
"wave_b_not_authorized",
]

View File

@@ -23,6 +23,7 @@ GRPC_BRIDGE = ROOT / "scripts/reboot-recovery/signoz-188-otlp-grpc-bridge.sh"
GRPC_UNIT = ROOT / "ops/systemd/user/awoooi-signoz-otlp-grpc-bridge.service"
GRPC_VERIFIER = ROOT / "scripts/reboot-recovery/verify-signoz-otel-grpc-runtime.sh"
GRPC_RUNTIME_RECEIPT = ROOT / "ops/signoz/p0-obs-002-grpc-runtime-receipt.yaml"
POST_CLOSURE_REGRESSION = ROOT / "ops/signoz/p0-obs-002-post-closure-regression.yaml"
def test_wave_a_contract_keeps_active_route_and_all_producers_on_incumbent() -> None:
@@ -266,6 +267,14 @@ def test_grpc_runtime_verifier_uses_real_service_aggregates_and_ten_minute_gate(
assert receipt["completion"]["grpc_runtime_asset"] == "closed"
assert receipt["completion"]["program"] == "partial_degraded"
regression = yaml.safe_load(POST_CLOSURE_REGRESSION.read_text(encoding="utf-8"))
assert regression["baseline"]["verifier_terminal"] == "pass"
assert regression["regressions"]["api_runtime"]["last_observed_cooldown_closed"] is False
assert regression["controlled_recovery"]["stateful_services_touched"] == 0
assert regression["controlled_recovery"]["grpc_4317_listening"] is True
assert regression["scope_boundaries"]["grpc_bridge_rollback_indicated"] is False
assert regression["terminal"]["status"] == "partial_degraded"
def test_managed_promotion_and_bridge_retirement_fail_closed_without_receipts() -> None:
env = os.environ.copy()