Files
osmedeus/internal/executor/bash_executor.go
T
j3ssie baac7a016a feat: add worker management, hooks support, and db cleanup enhancements
- Add worker eval command for distributed function execution with Redis hooks registration
- Add worker set command to update worker fields (alias, public-ip, ssh-enabled, ssh-keys-path)
- Enhance worker status with JSON output, search filtering, and column selection (--columns, --exclude-columns, --search)
- Add --keep-setting flag to install base/validate commands to preserve osm-settings.yaml after base installation
- Fix binary installation in Nix: replace CopyInstalledBinaryToFolder with SymlinkInstalledBinaryToFolder
- Add --clean-ws flag to db clean command for removing workspace data
- Add HooksEnabled field to Run records when creating runs from CLI and API
- Add comprehensive test coverage for hook execution (pre/post hooks, execution order, failure handling)
- Add test coverage for worker commands (eval, set, status with JSON) and db clean operations
- Improve usage documentation for worker subcommands and db operations
2026-02-15 10:47:44 +07:00

373 lines
11 KiB
Go

package executor
import (
"context"
"errors"
"fmt"
"os"
"os/exec"
"path/filepath"
"runtime"
"strings"
"sync"
"time"
"github.com/j3ssie/osmedeus/v5/internal/core"
"github.com/j3ssie/osmedeus/v5/internal/metrics"
"github.com/j3ssie/osmedeus/v5/internal/runner"
"github.com/j3ssie/osmedeus/v5/internal/template"
"go.uber.org/zap"
)
// BashExecutor executes bash steps
type BashExecutor struct {
templateEngine template.TemplateEngine
runner runner.Runner
}
// NewBashExecutor creates a new bash executor
func NewBashExecutor(engine template.TemplateEngine) *BashExecutor {
return &BashExecutor{
templateEngine: engine,
}
}
// Name returns the executor name for logging/debugging
func (e *BashExecutor) Name() string {
return "bash"
}
// StepTypes returns the step types this executor handles
func (e *BashExecutor) StepTypes() []core.StepType {
return []core.StepType{core.StepTypeBash}
}
// SetRunner sets the runner for command execution
func (e *BashExecutor) SetRunner(r runner.Runner) {
e.runner = r
}
// assembleCommand joins the command with structured args in order:
// command + speed_args + config_args + input_args + output_args
func assembleCommand(command, speedArgs, configArgs, inputArgs, outputArgs string) string {
parts := []string{command}
if speedArgs != "" {
parts = append(parts, speedArgs)
}
if configArgs != "" {
parts = append(parts, configArgs)
}
if inputArgs != "" {
parts = append(parts, inputArgs)
}
if outputArgs != "" {
parts = append(parts, outputArgs)
}
return strings.Join(parts, " ")
}
// writeStdFile writes command output to the specified file
func writeStdFile(path, content string) error {
// Ensure parent directory exists
if dir := filepath.Dir(path); dir != "" && dir != "." {
if err := os.MkdirAll(dir, 0755); err != nil {
return fmt.Errorf("failed to create directory: %w", err)
}
}
return os.WriteFile(path, []byte(content), 0644)
}
// extractToolName extracts the tool/binary name from a command string.
// Returns the base name of the first word in the command.
func extractToolName(command string) string {
parts := strings.Fields(command)
if len(parts) == 0 {
return "unknown"
}
return filepath.Base(parts[0])
}
// Execute executes a bash step
func (e *BashExecutor) Execute(ctx context.Context, step *core.Step, execCtx *core.ExecutionContext) (*core.StepResult, error) {
result := &core.StepResult{
StepName: step.Name,
Status: core.StepStatusRunning,
StartTime: time.Now(),
}
timeout, err := step.Timeout.Duration()
if err != nil {
result.Status = core.StepStatusFailed
result.Error = err
result.EndTime = time.Now()
result.Duration = result.EndTime.Sub(result.StartTime)
return result, err
}
// Extract binaries path for fallback resolution
binariesPath := ""
if bp, ok := execCtx.GetVariable("Binaries"); ok {
if bpStr, ok := bp.(string); ok {
binariesPath = bpStr
}
}
var output string
// Determine execution mode
if len(step.ParallelCommands) > 0 {
output, err = e.executeParallel(ctx, step.ParallelCommands, timeout, binariesPath)
} else if len(step.Commands) > 0 {
output, err = e.executeSequential(ctx, step.Commands, timeout, binariesPath)
} else if step.Command != "" {
// Assemble command with structured args if present
finalCmd := assembleCommand(step.Command, step.SpeedArgs, step.ConfigArgs, step.InputArgs, step.OutputArgs)
output, err = e.executeCommandWithFallback(ctx, finalCmd, timeout, binariesPath)
} else {
err = fmt.Errorf("no command specified")
}
result.Output = output
result.EndTime = time.Now()
result.Duration = result.EndTime.Sub(result.StartTime)
// Write stdout/stderr to file if std_file is specified
if step.StdFile != "" {
if writeErr := writeStdFile(step.StdFile, output); writeErr != nil {
// Log warning but don't fail the step
execCtx.Logger.Warn("Failed to write std_file",
zap.String("path", step.StdFile),
zap.Error(writeErr))
}
}
if err != nil {
result.Status = core.StepStatusFailed
result.Error = err
return result, err
}
result.Status = core.StepStatusSuccess
return result, nil
}
// executeCommand executes a single command
func (e *BashExecutor) executeCommand(ctx context.Context, command string, timeout time.Duration) (string, error) {
// Track execution timing for metrics
startTime := time.Now()
toolName := extractToolName(command)
// Apply timeout if specified
if timeout > 0 {
var cancel context.CancelFunc
ctx, cancel = context.WithTimeout(ctx, timeout)
defer cancel()
}
// Use runner if available, otherwise fall back to local execution
if e.runner != nil {
result, err := e.runner.Execute(ctx, command)
duration := time.Since(startTime).Seconds()
// Check for context cancellation first (Ctrl+C or timeout)
// This must be checked before other errors because runners return nil error
// but set result.ExitCode to -1 when context is cancelled
if ctx.Err() != nil {
if ctx.Err() == context.DeadlineExceeded {
metrics.RecordToolExecution(toolName, "timeout", duration)
return result.Output, fmt.Errorf("command timed out after %s", timeout)
}
metrics.RecordToolExecution(toolName, "cancelled", duration)
return result.Output, ctx.Err()
}
if err != nil {
metrics.RecordToolExecution(toolName, "error", duration)
return result.Output, fmt.Errorf("command failed: %w", err)
}
if result.ExitCode != 0 {
metrics.RecordToolExecution(toolName, "failed", duration)
return result.Output, newExitCodeErrorf(result.ExitCode, "command exited with code %d", result.ExitCode)
}
metrics.RecordToolExecution(toolName, "success", duration)
return strings.TrimSpace(result.Output), nil
}
// Fallback to local execution
// @NOTE: yes yes, I know this is a security risk. This is the intended behavior as it suppose to be that flexible so you should think twice of what you will run or design your workflow better
cmd := exec.CommandContext(ctx, "sh", "-c", command)
stdout := runner.NewLimitedBuffer(runner.MaxOutputSize)
stderr := runner.NewLimitedBuffer(runner.MaxStderrSize)
cmd.Stdout = stdout
cmd.Stderr = stderr
err := cmd.Run()
duration := time.Since(startTime).Seconds()
output := string(stdout.Bytes())
if stderr.Len() > 0 {
output += "\n" + string(stderr.Bytes())
}
if stdout.Overflow() || stderr.Overflow() {
output += "\n[output truncated]"
}
// Check for context cancellation first (Ctrl+C or timeout)
if ctx.Err() != nil {
if ctx.Err() == context.DeadlineExceeded {
metrics.RecordToolExecution(toolName, "timeout", duration)
return output, fmt.Errorf("command timed out after %s", timeout)
}
metrics.RecordToolExecution(toolName, "cancelled", duration)
return output, ctx.Err()
}
if err != nil {
// Extract exit code if available for fallback detection
var exitErr *exec.ExitError
if errors.As(err, &exitErr) {
exitCode := exitErr.ExitCode()
metrics.RecordToolExecution(toolName, "failed", duration)
return output, newExitCodeErrorf(exitCode, "command failed with exit code %d\nstderr: %s", exitCode, string(stderr.Bytes()))
}
metrics.RecordToolExecution(toolName, "error", duration)
return output, fmt.Errorf("command failed: %w\nstderr: %s", err, string(stderr.Bytes()))
}
metrics.RecordToolExecution(toolName, "success", duration)
return strings.TrimSpace(output), nil
}
// executeCommandWithFallback wraps executeCommand with automatic retry on exit code 127.
// Fallback 1: strip timeout prefix. Fallback 2: prepend binariesPath to the binary.
func (e *BashExecutor) executeCommandWithFallback(ctx context.Context, command string, timeout time.Duration, binariesPath string) (string, error) {
output, err := e.executeCommand(ctx, command, timeout)
if err == nil {
return output, nil
}
// Only attempt fallback on exit code 127 (command not found)
var ecErr *exitCodeError
if !errors.As(err, &ecErr) || ecErr.code != 127 {
return output, err
}
// Don't retry if context is already cancelled
if ctx.Err() != nil {
return output, err
}
currentCmd := command
// Fallback 1: strip timeout prefix
if result := stripTimeoutPrefix(currentCmd); result.stripped && result.command != "" {
// Use parsed duration from timeout prefix as fallback if step timeout is not set
retryTimeout := timeout
if retryTimeout == 0 && result.duration > 0 {
retryTimeout = result.duration
}
output, err = e.executeCommand(ctx, result.command, retryTimeout)
if err == nil {
return output, nil
}
// Check if still 127 for next fallback
if !errors.As(err, &ecErr) || ecErr.code != 127 {
return output, err
}
currentCmd = result.command
}
// Fallback 2: prepend binaries path
if prepended, ok := prependBinariesPath(currentCmd, binariesPath); ok {
output, err = e.executeCommand(ctx, prepended, timeout)
return output, err
}
return output, err
}
// executeSequential executes commands sequentially
func (e *BashExecutor) executeSequential(ctx context.Context, commands []string, timeout time.Duration, binariesPath string) (string, error) {
var outputs []string
for _, cmd := range commands {
output, err := e.executeCommandWithFallback(ctx, cmd, timeout, binariesPath)
outputs = append(outputs, output)
if err != nil {
return strings.Join(outputs, "\n"), err
}
}
return strings.Join(outputs, "\n"), nil
}
// executeParallel executes commands in parallel with bounded concurrency.
// Uses a worker pool capped at runtime.NumCPU()*2 to prevent unbounded goroutine/memory growth.
func (e *BashExecutor) executeParallel(ctx context.Context, commands []string, timeout time.Duration, binariesPath string) (string, error) {
type result struct {
index int
output string
err error
}
maxWorkers := runtime.NumCPU() * 2
if maxWorkers > len(commands) {
maxWorkers = len(commands)
}
type workItem struct {
index int
command string
}
workQueue := make(chan workItem, maxWorkers)
results := make(chan result, len(commands))
var wg sync.WaitGroup
// Start bounded worker pool
for range maxWorkers {
wg.Add(1)
go func() {
defer wg.Done()
for work := range workQueue {
output, err := e.executeCommandWithFallback(ctx, work.command, timeout, binariesPath)
results <- result{index: work.index, output: output, err: err}
}
}()
}
// Feed work items
go func() {
for i, cmd := range commands {
workQueue <- workItem{index: i, command: cmd}
}
close(workQueue)
}()
// Wait for all workers to complete
go func() {
wg.Wait()
close(results)
}()
// Collect results
outputs := make([]string, len(commands))
var firstError error
for r := range results {
outputs[r.index] = r.output
if r.err != nil && firstError == nil {
firstError = r.err
}
}
return strings.Join(outputs, "\n"), firstError
}
// CanHandle returns true if this executor can handle the given step type
func (e *BashExecutor) CanHandle(stepType core.StepType) bool {
return stepType == core.StepTypeBash
}