补齐人工处置PostgreSQL和MySQL实库执行器验证
This commit is contained in:
parent
0f2e6bfcce
commit
e0d16a6ab1
@ -0,0 +1,67 @@
|
||||
package database
|
||||
|
||||
import (
|
||||
"fmt"
|
||||
"testing"
|
||||
|
||||
"git.pcux.cn/xuans/Scert/backend/internal/testsupport/postgrestest"
|
||||
)
|
||||
|
||||
func TestPostgresRealRecoveryDispositionMigration(t *testing.T) {
|
||||
server := postgrestest.New(t)
|
||||
for version := 22; version <= 24; version++ {
|
||||
t.Run(fmt.Sprint(version), func(t *testing.T) {
|
||||
cfg := server.Database(t)
|
||||
db, err := Open(t.Context(), cfg)
|
||||
if err != nil {
|
||||
t.Fatal("隔离迁移连接失败")
|
||||
}
|
||||
t.Cleanup(func() { _ = db.Close() })
|
||||
if _, err := db.DB.ExecContext(t.Context(), migrationTableSQL(cfg.Driver)); err != nil {
|
||||
t.Fatal(err)
|
||||
}
|
||||
for previous := 1; previous < version; previous++ {
|
||||
if err := applyMigration(t.Context(), db.DB, cfg.Driver, previous); err != nil {
|
||||
t.Fatal(err)
|
||||
}
|
||||
}
|
||||
content, err := migrationFiles.ReadFile(fmt.Sprintf("migrations/postgres/%04d_%s.sql", version, migrationName(version)))
|
||||
if err != nil {
|
||||
t.Fatal(err)
|
||||
}
|
||||
// 已执行 DDL 但版本尚未落库时中止事务，不能留下半迁移结构。
|
||||
tx, err := db.DB.BeginTx(t.Context(), nil)
|
||||
if err != nil {
|
||||
t.Fatal(err)
|
||||
}
|
||||
for _, query := range migrationStatements(content) {
|
||||
if _, err := tx.ExecContext(t.Context(), query); err != nil {
|
||||
_ = tx.Rollback()
|
||||
t.Fatal(err)
|
||||
}
|
||||
}
|
||||
if err := tx.Rollback(); err != nil {
|
||||
t.Fatal(err)
|
||||
}
|
||||
current, err := currentMigrationVersion(t.Context(), db.DB)
|
||||
if err != nil || current != version-1 {
|
||||
t.Fatal("回滚后版本记录改变", err)
|
||||
}
|
||||
for range 2 {
|
||||
if err := Migrate(t.Context(), db.DB, cfg.Driver); err != nil {
|
||||
t.Fatal("中断后的迁移重入失败", err)
|
||||
}
|
||||
}
|
||||
current, err = currentMigrationVersion(t.Context(), db.DB)
|
||||
if err != nil || current != latestMigrationVersion {
|
||||
t.Fatal("迁移未达到当前版本", err)
|
||||
}
|
||||
for _, table := range []string{"recovery_dispositions", "recovery_claims", "recovery_events"} {
|
||||
var count int
|
||||
if err := db.DB.QueryRowContext(t.Context(), "SELECT COUNT(*) FROM "+table).Scan(&count); err != nil || count != 0 {
|
||||
t.Fatal("处置表未正确初始化", err)
|
||||
}
|
||||
}
|
||||
})
|
||||
}
|
||||
}
|
||||
212
backend/internal/repository/recovery_executor_external_test.go
Normal file
212
backend/internal/repository/recovery_executor_external_test.go
Normal file
@ -0,0 +1,212 @@
|
||||
package repository
|
||||
|
||||
import (
|
||||
"context"
|
||||
"database/sql"
|
||||
"errors"
|
||||
"maps"
|
||||
"reflect"
|
||||
"strings"
|
||||
"sync"
|
||||
"sync/atomic"
|
||||
"testing"
|
||||
|
||||
"github.com/go-sql-driver/mysql"
|
||||
|
||||
"git.pcux.cn/xuans/Scert/backend/internal/config"
|
||||
"git.pcux.cn/xuans/Scert/backend/internal/database"
|
||||
"git.pcux.cn/xuans/Scert/backend/internal/deployment"
|
||||
"git.pcux.cn/xuans/Scert/backend/internal/testsupport/mysqltest"
|
||||
"git.pcux.cn/xuans/Scert/backend/internal/testsupport/postgrestest"
|
||||
)
|
||||
|
||||
type recoveryExternalFixture struct {
|
||||
database func(*testing.T) config.DatabaseConfig
|
||||
inject func(*testing.T, *sql.DB, config.DatabaseConfig, string)
|
||||
}
|
||||
|
||||
func TestRecoveryExecutorPostgresReal(t *testing.T) {
|
||||
server := postgrestest.New(t)
|
||||
runExternalRecovery(t, recoveryExternalFixture{database: server.Database, inject: func(t *testing.T, db *sql.DB, _ config.DatabaseConfig, stage string) {
|
||||
for _, query := range []string{
|
||||
`CREATE FUNCTION reject_recovery_event() RETURNS trigger LANGUAGE plpgsql AS $$ BEGIN IF NEW.event_type = '` + stage + `' THEN RAISE EXCEPTION 'injected'; END IF; RETURN NEW; END $$`,
|
||||
`CREATE TRIGGER reject_recovery_event BEFORE INSERT ON recovery_events FOR EACH ROW EXECUTE FUNCTION reject_recovery_event()`,
|
||||
} {
|
||||
if _, err := db.ExecContext(t.Context(), query); err != nil {
|
||||
t.Fatal("安装隔离审计故障失败", err)
|
||||
}
|
||||
}
|
||||
}})
|
||||
}
|
||||
|
||||
func TestRecoveryExecutorMySQLReal(t *testing.T) {
|
||||
server := mysqltest.New(t)
|
||||
runExternalRecovery(t, recoveryExternalFixture{database: server.Database, inject: func(t *testing.T, _ *sql.DB, cfg config.DatabaseConfig, stage string) {
|
||||
parsed, err := mysql.ParseDSN(cfg.DSN)
|
||||
if err != nil || !strings.HasPrefix(parsed.DBName, "scert_test_") {
|
||||
t.Fatal("拒绝对非隔离库注入故障")
|
||||
}
|
||||
schema := "`" + strings.ReplaceAll(parsed.DBName, "`", "``") + "`"
|
||||
query := `CREATE TRIGGER ` + schema + `.reject_recovery_event BEFORE INSERT ON ` + schema + `.recovery_events FOR EACH ROW BEGIN IF NEW.event_type = '` + stage + `' THEN SIGNAL SQLSTATE '45000' SET MESSAGE_TEXT = 'injected'; END IF; END`
|
||||
// MySQL 开启二进制日志时，仅隔离管理连接创建触发器；业务账户不提权。
|
||||
if _, err := server.Admin().ExecContext(t.Context(), query); err != nil {
|
||||
t.Fatal("安装隔离审计故障失败", err)
|
||||
}
|
||||
}})
|
||||
}
|
||||
|
||||
func openRecoveryExternal(t *testing.T, cfg config.DatabaseConfig) *sql.DB {
|
||||
t.Helper()
|
||||
db, err := database.Open(t.Context(), cfg)
|
||||
if err != nil {
|
||||
t.Fatal("连接隔离实库失败")
|
||||
}
|
||||
t.Cleanup(func() { _ = db.Close() })
|
||||
return db.DB
|
||||
}
|
||||
|
||||
// 只替换远端文件实现，准备、确认、执行、迁移、占用和审计均走实际产品代码。
|
||||
func runExternalRecovery(t *testing.T, fixture recoveryExternalFixture) {
|
||||
t.Helper()
|
||||
for _, action := range []deployment.RecoveryAction{deployment.RecoveryActionRecord, deployment.RecoveryActionCommit, deployment.RecoveryActionRollback} {
|
||||
for _, stage := range []string{"none", "started", "finished"} {
|
||||
t.Run(string(action)+"/"+stage, func(t *testing.T) {
|
||||
cfg := fixture.database(t)
|
||||
db := openRecoveryExternal(t, cfg)
|
||||
if err := database.Migrate(t.Context(), db, cfg.Driver); err != nil {
|
||||
t.Fatal(err)
|
||||
}
|
||||
repo, _, service, client, server, record := seedRecoveryExecutor(t, NewSQLDeploymentTransactionRepository(db, cfg.Driver), db)
|
||||
before, err := repo.ListPendingDeploymentTransactions(t.Context())
|
||||
if err != nil {
|
||||
t.Fatal(err)
|
||||
}
|
||||
filesBefore := maps.Clone(client.files)
|
||||
resolve := func(context.Context, string) (deployment.Server, error) { return server, nil }
|
||||
confirmation, token, err := service.PrepareRecoveryConfirmation(t.Context(), record.ID, "unix:0", action, resolve)
|
||||
if err != nil {
|
||||
t.Fatal("实库准备失败", err)
|
||||
}
|
||||
if stage != "none" {
|
||||
fixture.inject(t, db, cfg, stage)
|
||||
}
|
||||
item, err := service.ExecuteRecoveryDisposition(t.Context(), confirmation, "unix:0", token, resolve)
|
||||
if stage == "none" {
|
||||
if err != nil || item.Status != "succeeded" {
|
||||
t.Fatal("实库执行未成功", err)
|
||||
}
|
||||
recoveryCount(t, db, "recovery_dispositions", 1)
|
||||
recoveryCount(t, db, "recovery_claims", 0)
|
||||
recoveryCount(t, db, "recovery_events", 2)
|
||||
} else if !errors.Is(err, deployment.ErrRecoveryDispositionUnavailable) || item.Status == "succeeded" {
|
||||
t.Fatal("审计故障未拒绝成功结果", err)
|
||||
}
|
||||
if _, err := service.ExecuteRecoveryDisposition(t.Context(), confirmation, "unix:0", token, resolve); !errors.Is(err, deployment.ErrRecoveryConfirmation) {
|
||||
t.Fatal("实库执行确认被重放", err)
|
||||
}
|
||||
// 实际关闭连接池后重新打开，不能靠旧仓储对象的内存证明持久性。
|
||||
if err := db.Close(); err != nil {
|
||||
t.Fatal(err)
|
||||
}
|
||||
db = openRecoveryExternal(t, cfg)
|
||||
if err := database.Migrate(t.Context(), db, cfg.Driver); err != nil {
|
||||
t.Fatal("实库重开迁移失败", err)
|
||||
}
|
||||
repo = NewSQLDeploymentTransactionRepository(db, cfg.Driver)
|
||||
pending, err := repo.ListPendingDeploymentTransactions(t.Context())
|
||||
if err != nil {
|
||||
t.Fatal(err)
|
||||
}
|
||||
if stage == "started" {
|
||||
for _, table := range []string{"recovery_dispositions", "recovery_claims", "recovery_events"} {
|
||||
recoveryCount(t, db, table, 0)
|
||||
}
|
||||
if !reflect.DeepEqual(before, pending) || !reflect.DeepEqual(filesBefore, client.files) {
|
||||
t.Fatal("开始审计失败修改了原事务或远端")
|
||||
}
|
||||
return
|
||||
}
|
||||
stored, err := repo.GetRecoveryDisposition(t.Context(), confirmation.DispositionID())
|
||||
if err != nil {
|
||||
t.Fatal("重开后处置记录丢失", err)
|
||||
}
|
||||
restarted := deployment.NewServiceWithJournal(nil, repo)
|
||||
view, err := restarted.GetRecoveryDisposition(t.Context(), confirmation.DispositionID())
|
||||
if err != nil || view.Validate() != nil || !reflect.DeepEqual(view, stored.Summary()) {
|
||||
t.Fatal("重开后脱敏查询与持久结果不一致", err)
|
||||
}
|
||||
if stage == "finished" {
|
||||
recoveryCount(t, db, "recovery_claims", 1)
|
||||
recoveryCount(t, db, "recovery_events", 1)
|
||||
if stored.Status != "running" || len(pending) != 1 || pending[0].Status != deployment.TransactionStatusRecoveryRequired {
|
||||
t.Fatal("结束审计失败丢失占用或发布了事务终态")
|
||||
}
|
||||
if err := restarted.RecoverPending(t.Context(), func(context.Context, string) (deployment.Server, error) {
|
||||
t.Error("重开后自动恢复重复执行人工占用事务")
|
||||
return deployment.Server{}, errors.New("禁止解析远端")
|
||||
}); !errors.Is(err, deployment.ErrRecoveryDispositionBusy) {
|
||||
t.Fatal("重开后未遵守持久占用", err)
|
||||
}
|
||||
} else if stored.Status != "succeeded" {
|
||||
t.Fatal("成功结果重开后丢失")
|
||||
}
|
||||
if action == deployment.RecoveryActionRecord {
|
||||
if !reflect.DeepEqual(before, pending) || !reflect.DeepEqual(filesBefore, client.files) {
|
||||
t.Fatal("仅记录动作改变原事务、证据或远端文件")
|
||||
}
|
||||
} else {
|
||||
expected, result := "new", deployment.RecoveryResultCommitted
|
||||
if action == deployment.RecoveryActionRollback {
|
||||
expected, result = "old", deployment.RecoveryResultRolledBack
|
||||
}
|
||||
for _, path := range record.Paths {
|
||||
if client.files[path] != expected {
|
||||
t.Fatal("远端动作与明确选择不一致")
|
||||
}
|
||||
}
|
||||
if stage == "none" && (len(pending) != 0 || stored.ResultCode != result) {
|
||||
t.Fatal("成功结果与原事务终态不一致")
|
||||
}
|
||||
}
|
||||
})
|
||||
}
|
||||
}
|
||||
t.Run("concurrent_idempotency", func(t *testing.T) {
|
||||
cfg := fixture.database(t)
|
||||
db := openRecoveryExternal(t, cfg)
|
||||
if err := database.Migrate(t.Context(), db, cfg.Driver); err != nil {
|
||||
t.Fatal(err)
|
||||
}
|
||||
repo, _, intent := seedRecoveryDisposition(t, NewSQLDeploymentTransactionRepository(db, cfg.Driver), db, deployment.RecoveryActionRecord)
|
||||
var created atomic.Int32
|
||||
var group sync.WaitGroup
|
||||
for range 8 {
|
||||
group.Go(func() {
|
||||
_, fresh, err := repo.BeginRecoveryDisposition(t.Context(), intent)
|
||||
if err != nil {
|
||||
t.Errorf("并发幂等请求失败：%v", err)
|
||||
}
|
||||
if fresh {
|
||||
created.Add(1)
|
||||
}
|
||||
})
|
||||
}
|
||||
group.Wait()
|
||||
if created.Load() != 1 {
|
||||
t.Fatal("并发请求没有精确创建一次处置")
|
||||
}
|
||||
conflict := intent
|
||||
conflict.Operator = "unix:1"
|
||||
if _, _, err := repo.BeginRecoveryDisposition(t.Context(), conflict); !errors.Is(err, deployment.ErrRecoveryDispositionConflict) {
|
||||
t.Fatal("不同操作者复用了幂等结果", err)
|
||||
}
|
||||
conflict = intent
|
||||
conflict.ID, conflict.RequestKey = strings.Repeat("e", 64), strings.Repeat("f", 64)
|
||||
if _, _, err := repo.BeginRecoveryDisposition(t.Context(), conflict); !errors.Is(err, deployment.ErrRecoveryDispositionBusy) {
|
||||
t.Fatal("其他请求覆盖已有占用", err)
|
||||
}
|
||||
for _, table := range []string{"recovery_dispositions", "recovery_claims", "recovery_events"} {
|
||||
recoveryCount(t, db, table, 1)
|
||||
}
|
||||
})
|
||||
}
|
||||
@ -77,6 +77,11 @@ func executorIntegrationFixture(t *testing.T) (*SQLDeploymentTransactionReposito
|
||||
_, db := newTaskRepositoryTest(t)
|
||||
t.Cleanup(func() { _ = db.Close() })
|
||||
repo := NewSQLDeploymentTransactionRepository(db)
|
||||
return seedRecoveryExecutor(t, repo, db)
|
||||
}
|
||||
|
||||
func seedRecoveryExecutor(t *testing.T, repo *SQLDeploymentTransactionRepository, db *sql.DB) (*SQLDeploymentTransactionRepository, *sql.DB, *deployment.Service, *executorIntegrationClient, deployment.Server, deployment.TransactionRecord) {
|
||||
t.Helper()
|
||||
server := deployment.Server{
|
||||
ID: "executor-server", Name: "executor-server", Address: "192.0.2.10", Port: 22, Username: "deploy",
|
||||
AuthMethod: deployment.SSHAuthPassword, PasswordRef: "secret://ssh/password/test", HostKeyFingerprint: "SHA256:test",
|
||||
|
||||
197
backend/internal/testsupport/postgrestest/server.go
Normal file
197
backend/internal/testsupport/postgrestest/server.go
Normal file
@ -0,0 +1,197 @@
|
||||
package postgrestest
|
||||
|
||||
import (
|
||||
"context"
|
||||
"crypto/ecdsa"
|
||||
"crypto/elliptic"
|
||||
"crypto/rand"
|
||||
"crypto/x509"
|
||||
"crypto/x509/pkix"
|
||||
"database/sql"
|
||||
"encoding/hex"
|
||||
"encoding/pem"
|
||||
"fmt"
|
||||
"math/big"
|
||||
"net"
|
||||
"net/url"
|
||||
"os"
|
||||
"os/exec"
|
||||
"path/filepath"
|
||||
"strconv"
|
||||
"strings"
|
||||
"testing"
|
||||
"time"
|
||||
|
||||
_ "github.com/jackc/pgx/v5/stdlib"
|
||||
|
||||
"git.pcux.cn/xuans/Scert/backend/internal/config"
|
||||
)
|
||||
|
||||
// Server 只管理本次测试创建的隔离集群，不接受已有实例的连接串。
|
||||
type Server struct {
|
||||
root string
|
||||
bin string
|
||||
port string
|
||||
password string
|
||||
admin *sql.DB
|
||||
sequence int
|
||||
}
|
||||
|
||||
func New(t *testing.T) *Server {
|
||||
t.Helper()
|
||||
bin, root := os.Getenv("SCERT_POSTGRES_TEST_BIN"), os.Getenv("SCERT_POSTGRES_TEST_ROOT")
|
||||
if bin == "" {
|
||||
t.Skip("未启用 PostgreSQL 隔离实库测试")
|
||||
}
|
||||
if !filepath.IsAbs(bin) || !filepath.IsAbs(root) {
|
||||
t.Fatal("PostgreSQL 工具与隔离产物目录必须使用绝对路径")
|
||||
}
|
||||
if err := os.MkdirAll(root, 0o700); err != nil {
|
||||
t.Fatal(err)
|
||||
}
|
||||
root, err := os.MkdirTemp(root, "scert-postgres-integration-")
|
||||
if err != nil {
|
||||
t.Fatal(err)
|
||||
}
|
||||
var secret [32]byte
|
||||
if _, err := rand.Read(secret[:]); err != nil {
|
||||
t.Fatal(err)
|
||||
}
|
||||
f := &Server{root: root, bin: bin, password: hex.EncodeToString(secret[:])}
|
||||
t.Logf("PostgreSQL 隔离产物保留于：%s", root)
|
||||
ctx, cancel := context.WithTimeout(t.Context(), 10*time.Second)
|
||||
defer cancel()
|
||||
cmd := exec.CommandContext(ctx, filepath.Join(bin, "postgres"), "--version")
|
||||
cmd.Env = f.environment()
|
||||
version, err := cmd.Output()
|
||||
if err != nil || !strings.HasPrefix(string(version), "postgres (PostgreSQL) ") {
|
||||
t.Fatal("无法确认隔离 PostgreSQL 版本")
|
||||
}
|
||||
t.Logf("隔离服务端：%s", strings.TrimSpace(string(version)))
|
||||
passwordFile := filepath.Join(root, "bootstrap-password.txt")
|
||||
write(t, passwordFile, []byte(f.password+"\n"))
|
||||
data := filepath.Join(root, "cluster")
|
||||
f.control(t, "initdb", "-D", data, "-U", "postgres", "--pwfile="+passwordFile, "--auth=scram-sha-256", "--encoding=UTF8", "--no-locale")
|
||||
f.createTLS(t)
|
||||
listener, err := net.Listen("tcp4", "127.0.0.1:0")
|
||||
if err != nil {
|
||||
t.Fatal(err)
|
||||
}
|
||||
f.port = strconv.Itoa(listener.Addr().(*net.TCPAddr).Port)
|
||||
if err := listener.Close(); err != nil {
|
||||
t.Fatal(err)
|
||||
}
|
||||
settings := "shared_buffers = '16MB'\nwork_mem = '1MB'\nmaintenance_work_mem = '16MB'\nmax_connections = 24\nmax_parallel_workers = 0\nmax_worker_processes = 0\n"
|
||||
settings += "listen_addresses = '127.0.0.1'\nport = " + f.port + "\nssl = on\nssl_cert_file = 'server.crt'\nssl_key_file = 'server.key'\npassword_encryption = 'scram-sha-256'\nunix_socket_directories = ''\n"
|
||||
write(t, filepath.Join(data, "postgresql.conf"), []byte(settings))
|
||||
write(t, filepath.Join(data, "pg_hba.conf"), []byte("hostssl all all 127.0.0.1/32 scram-sha-256\nhost all all 0.0.0.0/0 reject\nhost all all ::0/0 reject\n"))
|
||||
// 即使启动中途失败，也只停止本测试目录中的集群；不删除产物。
|
||||
t.Cleanup(func() {
|
||||
if f.admin != nil {
|
||||
_ = f.admin.Close()
|
||||
}
|
||||
if _, err := os.Stat(filepath.Join(data, "postmaster.pid")); os.IsNotExist(err) {
|
||||
return
|
||||
}
|
||||
ctx, cancel := context.WithTimeout(context.Background(), 30*time.Second)
|
||||
defer cancel()
|
||||
cmd := exec.CommandContext(ctx, filepath.Join(bin, "pg_ctl"), "-D", data, "-m", "fast", "-w", "-t", "20", "stop")
|
||||
cmd.Env = f.environment()
|
||||
cmd.WaitDelay = 2 * time.Second
|
||||
if err := cmd.Run(); err != nil {
|
||||
t.Errorf("停止隔离 PostgreSQL 失败：%v", err)
|
||||
}
|
||||
})
|
||||
f.control(t, "pg_ctl", "-D", data, "-l", filepath.Join(root, "postgres.log"), "-w", "-t", "30", "start")
|
||||
f.admin, err = sql.Open("pgx", f.dsn("postgres", "postgres"))
|
||||
if err != nil || f.admin.PingContext(t.Context()) != nil {
|
||||
t.Fatal("隔离 PostgreSQL 严格 TLS 管理连接失败")
|
||||
}
|
||||
return f
|
||||
}
|
||||
|
||||
func (f *Server) Database(t *testing.T) config.DatabaseConfig {
|
||||
t.Helper()
|
||||
f.sequence++
|
||||
name, owner := fmt.Sprintf("scert_test_%d", f.sequence), fmt.Sprintf("scert_owner_%d", f.sequence)
|
||||
// 每个用例都有独立数据库和无超级权限的属主，不能访问业务数据库。
|
||||
for _, query := range []string{
|
||||
"CREATE ROLE " + owner + " LOGIN NOSUPERUSER NOCREATEDB NOCREATEROLE PASSWORD '" + f.password + "'",
|
||||
"CREATE DATABASE " + name + " OWNER " + owner,
|
||||
"REVOKE ALL ON DATABASE " + name + " FROM PUBLIC",
|
||||
} {
|
||||
if _, err := f.admin.ExecContext(t.Context(), query); err != nil {
|
||||
t.Fatal("创建隔离 PostgreSQL 数据库及受限属主失败")
|
||||
}
|
||||
}
|
||||
return config.DatabaseConfig{Driver: "postgres", DSN: f.dsn(name, owner), MaxOpenConns: 8, MaxIdleConns: 2}
|
||||
}
|
||||
|
||||
func (f *Server) dsn(name, user string) string {
|
||||
u := &url.URL{Scheme: "postgres", Host: net.JoinHostPort("127.0.0.1", f.port), Path: "/" + name, User: url.UserPassword(user, f.password)}
|
||||
u.RawQuery = url.Values{"sslmode": {"verify-full"}, "sslrootcert": {filepath.Join(f.root, "cluster", "server.crt")}, "connect_timeout": {"5"}}.Encode()
|
||||
return u.String()
|
||||
}
|
||||
|
||||
func (f *Server) environment() []string {
|
||||
env := []string{}
|
||||
for _, entry := range os.Environ() {
|
||||
key, _, _ := strings.Cut(entry, "=")
|
||||
upper := strings.ToUpper(key)
|
||||
if strings.HasPrefix(upper, "PG") || upper == "PATH" || upper == "LC_ALL" || upper == "LANG" {
|
||||
continue
|
||||
}
|
||||
env = append(env, entry)
|
||||
}
|
||||
return append(env, "PATH="+f.bin+string(os.PathListSeparator)+os.Getenv("PATH"), "LC_ALL=C", "LANG=C")
|
||||
}
|
||||
|
||||
func (f *Server) control(t *testing.T, tool string, args ...string) {
|
||||
t.Helper()
|
||||
ctx, cancel := context.WithTimeout(t.Context(), time.Minute)
|
||||
defer cancel()
|
||||
cmd := exec.CommandContext(ctx, filepath.Join(f.bin, tool), args...)
|
||||
cmd.Env = f.environment()
|
||||
// 后台进程不能继承 Go 输出复制管道，否则父进程会一直等待。
|
||||
output, err := os.OpenFile(filepath.Join(f.root, tool+".log"), os.O_CREATE|os.O_APPEND|os.O_WRONLY, 0o600)
|
||||
if err != nil {
|
||||
t.Fatal(err)
|
||||
}
|
||||
defer output.Close()
|
||||
cmd.Stdout, cmd.Stderr = output, output
|
||||
cmd.WaitDelay = 2 * time.Second
|
||||
if err := cmd.Run(); err != nil {
|
||||
t.Fatalf("隔离工具 %s 失败，请检查保留日志：%v", tool, err)
|
||||
}
|
||||
}
|
||||
|
||||
func write(t *testing.T, path string, value []byte) {
|
||||
t.Helper()
|
||||
if err := os.WriteFile(path, value, 0o600); err != nil {
|
||||
t.Fatal(err)
|
||||
}
|
||||
}
|
||||
|
||||
func (f *Server) createTLS(t *testing.T) {
|
||||
t.Helper()
|
||||
key, err := ecdsa.GenerateKey(elliptic.P256(), rand.Reader)
|
||||
if err != nil {
|
||||
t.Fatal(err)
|
||||
}
|
||||
certificate := &x509.Certificate{
|
||||
SerialNumber: big.NewInt(1), Subject: pkix.Name{CommonName: "Scert isolated PostgreSQL"},
|
||||
NotBefore: time.Now().Add(-time.Hour), NotAfter: time.Now().Add(24 * time.Hour),
|
||||
IPAddresses: []net.IP{net.ParseIP("127.0.0.1")}, IsCA: true, BasicConstraintsValid: true,
|
||||
KeyUsage: x509.KeyUsageCertSign | x509.KeyUsageDigitalSignature, ExtKeyUsage: []x509.ExtKeyUsage{x509.ExtKeyUsageServerAuth},
|
||||
}
|
||||
der, err := x509.CreateCertificate(rand.Reader, certificate, certificate, &key.PublicKey, key)
|
||||
if err != nil {
|
||||
t.Fatal(err)
|
||||
}
|
||||
private, err := x509.MarshalPKCS8PrivateKey(key)
|
||||
if err != nil {
|
||||
t.Fatal(err)
|
||||
}
|
||||
write(t, filepath.Join(f.root, "cluster", "server.crt"), pem.EncodeToMemory(&pem.Block{Type: "CERTIFICATE", Bytes: der}))
|
||||
write(t, filepath.Join(f.root, "cluster", "server.key"), pem.EncodeToMemory(&pem.Block{Type: "PRIVATE KEY", Bytes: private}))
|
||||
}
|
||||
@ -1,5 +1,17 @@
|
||||
# Linux 生产流程测试记录
|
||||
|
||||
## 2026年10月3日21:48 外部数据库专项
|
||||
|
||||
- [✓] 21:41 实查 `scert-local-actions-fixed-20261003-2125.service` 已 inactive、MainPID=0、Result=success，`LOCAL_ACTIONS_FIXED_HTTP_FULL_RACE_EXIT=0`；HTTP 全包竞态约 393.752 秒，五项检查完整通过。
|
||||
- [x] 新四份测试源码归档 SHA-256 为 `421b52f1916f7be8cf088e0c0bd9623ad71e777f7b1439277a7e58faa4045e24`，本地与第一台一致；旧任务终止后才更新隔离源码。
|
||||
- [x] PostgreSQL 14.24 工具经 APT 下载校验后仅解压在 `/var/tmp/scert-pg14-disposition-20261003`，未安装系统服务。Git 隔离目录属主检查首次阻止任务启动，改用单条命令的 safe.directory 后启动，没有修改全局 Git 配置或重复启动测试。
|
||||
- [✓] `scert-external-disposition-20261003-2143.service` 的 `external-disposition.log`：PostgreSQL 迁移 22—24 三场景通过，数据库包约 3.672 秒；执行器十场景通过，仓储包约 7.603 秒；`EXTERNAL_POSTGRES_EXIT=0`。
|
||||
- [✓] 同日志 MySQL 5.7.44 执行器十场景通过，约 25.574 秒；`EXTERNAL_MYSQL_EXIT=0`。三动作分别覆盖正常、开始审计失败、结束审计失败，另外包含八并发幂等；重开的是实际连接池，远端文件和站点核验是测试实现。
|
||||
- [x] 21:46 查询无 PostgreSQL 残留进程，测试自动停止独立集群；仅 SQL 存储专项，不替代 PostgreSQL 16 备份工具、正式包或首次 Web 验收。
|
||||
- [✓] 本地 `npm.cmd test`：275 项通过，零失败、零跳过。
|
||||
- [~] 最新 SQLite 执行器、完整静态检查及后端普通全量仍在上述同一单元继续执行；不提前标记整组成功。
|
||||
|
||||
|
||||
## 2026年10月3日21:32 本机接入修复版回归
|
||||
|
||||
- [x] 首轮任务 `scert-local-actions-20261003-2116.service` 已失败终止，`LOCAL_ACTIONS_RACE_EXIT=1`。分片测试夹具使用默认四 KiB 读取器导致过早失败；已改用协议最大帧大小的读取器，没有放宽生产协议。HTTP 全包竞态五分钟超时，堆栈位于既有 SQLite 迁移，不能记为通过。
|
||||
|
||||
@ -1,5 +1,13 @@
|
||||
# Scert 开发任务清单
|
||||
|
||||
## 2026年10月3日21:48 当前开发进度
|
||||
|
||||
- [✓] 本机协议/终端接入的五项 Linux 回归全部正常结束；`0f2e6bf` 已提交推送。
|
||||
- [x] 新增隔离 PostgreSQL 测试支持和跨 PostgreSQL/MySQL 的真实执行器矩阵，复用远端夹具，数据库和服务执行链均使用实际产品代码。
|
||||
- [✓] PostgreSQL 迁移三场景、执行器十场景及 MySQL 执行器十场景通过；含审计失败、真实连接池重开、占用和并发幂等。
|
||||
- [✓] 前端 275 项测试通过；最新后端剩余回归继续运行，正式包与生产闭环尚未完成。
|
||||
|
||||
|
||||
## 2026年10月3日21:32 当前开发进度
|
||||
|
||||
- [x] 受保护本机会话、协议和终端三动作及结果查询已编码；确认绑定会话和真实对端，一次消费，失败或断连不重放。
|
||||
|
||||
@ -1,12 +1,21 @@
|
||||
# Scert 整改任务执行清单
|
||||
|
||||
## 2026年10月3日21:48 最新执行结论
|
||||
|
||||
- [✓] 本机接入整组 Linux 回归正常结束，五项退出码零；检查点 `0f2e6bf` 已推送。
|
||||
- [✓] PostgreSQL 14.24 迁移三场景、执行器十场景和 MySQL 5.7.44 执行器十场景通过；仅隔离实库，不是实际 SSH/HTTPS 验收。
|
||||
- [✓] 前端 275 项测试通过；第二台磁盘剩余 26 GB，不再作为阻塞。
|
||||
- [~] 最新 SQLite 执行器、静态检查和后端全量仍运行；之后构建正式 Linux 包。
|
||||
- [ ] 三数据库首次 Web、系统邮件和双机生产门禁仍待完成。
|
||||
|
||||
|
||||
## 2026年10月3日21:32 最新执行结论
|
||||
|
||||
- [x] R3-F/G/H 本机会话、协议隔离、终端风险确认及断连查询已接入。
|
||||
- [✓] 修复版 Linux 三包竞态、HTTP 本机专项竞态、静态检查和后端全量退出码零。
|
||||
- [~] HTTP 全包竞态仍运行，不能把运行中的整组标通过；首次夹具失败和 HTTP 五分钟超时不隐去。
|
||||
- [~] 继续 R4 必要外部数据库处置原子性与重开查询；后续为 R5 正式 Linux 包与双机验收。
|
||||
- [!] 第二台磁盘已实查剩余 26 GB，不再属于阻塞；系统邮件和完整生产门禁仍未通过。
|
||||
- [x] 第二台磁盘已实查剩余 26 GB，不再属于阻塞；系统邮件和完整生产门禁仍未通过。
|
||||
|
||||
|
||||
## 2026年10月3日21:00 最新执行结论
|
||||
|
||||
@ -1,5 +1,15 @@
|
||||
# Scert 整改、合并与优化清单
|
||||
|
||||
## 2026年10月3日21:48 实库与本机入口进度
|
||||
|
||||
- [✓] R3 本机接入修复版任务已正常结束，五项退出码零，HTTP 全包竞态约 393.752 秒；代码检查点 `0f2e6bf` 已推送。
|
||||
- [✓] R4 PostgreSQL 14.24 隔离实库：迁移 22—24 三个事务回滚重入场景，以及执行器三动作、两类审计故障、重开查询/占用、八并发幂等共十个场景通过。
|
||||
- [✓] R4 MySQL 5.7.44 同一执行器十场景通过，业务连接仍为受限账户；只在本次隔离数据库注入审计故障。
|
||||
- [~] 最新 SQLite 执行器竞态、静态检查和普通后端全量正在继续；外部实库证明不包括实际远端文件或 TLS 站点动作。
|
||||
- [✓] 当前前端 275 项测试全部通过，无失败和跳过。
|
||||
- [ ] 下一项为正式 Linux 候选包和双机生产验证；PostgreSQL 本轮为 SQL 持久化专项，不代表 PostgreSQL 16 备份工具、首次 Web 或最终生产验收。
|
||||
|
||||
|
||||
## 2026年10月3日21:32 最新整改进度
|
||||
|
||||
本节优先于历史状态；仅 Linux，生产验收仍未通过。
|
||||
|
||||
@ -1,5 +1,14 @@
|
||||
# 部署事务人工处置设计
|
||||
|
||||
## 2026年10月3日21:48 外部持久化证据
|
||||
|
||||
- [✓] 本机接入五项回归已全部通过，包括 HTTP 全包竞态。
|
||||
- [✓] 真实 PostgreSQL 14.24 和 MySQL 5.7.44 的产品执行器通过三动作、开始与结束审计故障、重开查询/占用和并发幂等；远端文件仍为测试实现。
|
||||
- [✓] PostgreSQL 新迁移 22—24 在事务 DDL 执行后回滚，再次迁移可达到连续版本 24，并可重复迁移。
|
||||
- [x] PostgreSQL 仅回环且 verify-full，临时集群随机凭据，业务连接不是超级用户；测试不接受现有实例连接串，不清理业务数据库。
|
||||
- [ ] 正式安装包与真实 SSH 文件及 HTTPS 站点闭环仍需直接验证。
|
||||
|
||||
|
||||
## 2026年10月3日21:32 本机接入实现
|
||||
|
||||
- [x] 准备确认对象仅留在服务端会话，终端只取得短期令牌和脱敏摘要；执行先清空确认，所有失败尝试均不能再次使用。
|
||||
|
||||
@ -1,5 +1,14 @@
|
||||
# 部署事务人工核查任务清单
|
||||
|
||||
## 2026年10月3日21:48 外部数据库复核
|
||||
|
||||
- [✓] R3-F/G/H 本机接入源码回归全部结束，HTTP 全包竞态通过；正式包入口尚待验证。
|
||||
- [✓] R4-D PostgreSQL 14.24：迁移 22—24 中止事务后重入通过；三动作、开始/结束审计失败、连接池重开查询与占用，以及八并发幂等通过。
|
||||
- [✓] R4-E MySQL 5.7.44：执行器同一十场景通过，不提升业务账户权限。
|
||||
- [~] 新增测试后的 SQLite 执行器回归、静态检查与后端全量正在执行。
|
||||
- [ ] R5 正式包、实际终端与双机文件/HTTPS 动作，以及完整首次 Web 与系统邮件门禁尚未完成。
|
||||
|
||||
|
||||
## 2026年10月3日21:32 本机接入与下一批任务
|
||||
|
||||
沿用已确认需求，不增加恢复网络端口，不处理 Windows。
|
||||
|
||||
Loading…
Reference in New Issue
Block a user