Files
orca/internal/cli/job.go
T
Jon Chery 64cbbd543e feat(C): remote deployment correctness — PVE runtime, DB record, Traefik (REQ-166)
- deployRemote branches on runtime: pve-ct/pve-vm on proxmox nodes
  invoke runtime.Registry.Prepare+Start (creates LXC/VM via SSH);
  process runtime on linux nodes uses systemd emitter; process on
  proxmox is rejected with clear error
- deployRemote emits Traefik dynamic config when spec has ports
  (TraefikEmitter.Render + SSH-push to /etc/traefik/dynamic/)
- model.Job: added Node field so job list --json reports deployed node
- job run remote case: inserts model.Job + alloc_history after
  deployRemote succeeds (job list and job stop now work for remote)
- --target name lookup: legacy dispatcher tries NodeID match
- job list UX: short 8-char IDs, NODE column, conditional EXIT (- for
  non-terminal statuses)
- emitter/traefik.go: directory provider (was single file), pve-ct/pve-vm
  registered in RegisterTraefik
- runtime/pve.go: shellQuote for image, idempotent create check

---ci---
project: orca
milestone: v0.12.18
phase: C
status: complete
requirements:
  covered: [166]
---/ci---
2026-08-10 16:21:50 +00:00

620 lines
19 KiB
Go

package cli
import (
"context"
"database/sql"
"encoding/json"
"errors"
"fmt"
"os"
"os/signal"
"strings"
"syscall"
"time"
"github.com/google/uuid"
"github.com/spf13/cobra"
"git.cloudinit.dev/coreci/orca/internal/certpaths"
"git.cloudinit.dev/coreci/orca/internal/engine"
"git.cloudinit.dev/coreci/orca/internal/jobspec"
"git.cloudinit.dev/coreci/orca/internal/model"
"git.cloudinit.dev/coreci/orca/internal/sshpush"
"git.cloudinit.dev/coreci/orca/internal/store"
)
var jobCmd = &cobra.Command{
Use: "job",
Short: "Manage orca jobs",
Long: "Run, list, stop, and inspect orca jobs.",
}
func jobExecutor() (*engine.Executor, func() error, error) {
db, closer, err := openDB()
if err != nil {
return nil, nil, err
}
jobs := store.NewJobRepo(db)
tasks := store.NewTaskRepo(db)
return engine.NewExecutor(jobs, tasks, newLogger()), closer, nil
}
var (
stopID string
runTarget string
runIDKey string
jobWatch bool
)
var jobRunCmd = &cobra.Command{
Use: "run <spec.md>",
Short: "Run a job from a markdown spec file",
Long: "Submit a job spec, execute its tasks, and persist the result. Use --target to pin to a specific node (overrides bin-packing); --idempotency-key for cross-node dispatch dedupe.",
Args: cobra.ExactArgs(1),
RunE: func(cmd *cobra.Command, args []string) error {
if strings.HasSuffix(args[0], ".hcl") {
warnDeprecated("orca job run <spec.hcl> is deprecated: .hcl jobspec is legacy (R-013); convert to .md format (REQ-064) — see .ciagent/PRD_v0.9.md")
}
spec, err := jobspec.ParseFile(args[0])
if err != nil {
return err
}
ctx, cancel := context.WithTimeout(cmd.Context(), 5*time.Minute)
defer cancel()
// v0.13 phase-03 scheduler wiring (REQ-151, C-44): decide
// whether to run locally (dev mode / no remote nodes) or
// remotely (scheduler picks a peer, render systemd, SSH-push).
// The deprecated mTLS Dispatcher path (--idempotency-key) is
// retained only for the dual-write window; the new remote path
// uses the CLI-side scheduler + sshpush.
if runIDKey != "" {
// Legacy --idempotency-key dispatch path (deprecated mTLS
// Dispatcher). Retained for backward compat; routes through
// engine.Dispatcher which is scheduled for removal in v0.10.
exec, closer, err := jobExecutor()
if err != nil {
return err
}
defer closer()
db, dbCloser, err := openDB()
if err != nil {
return err
}
defer dbCloser()
peers := engine.NewPeerRegistry()
dispatcher := engine.NewDispatcher(newLogger(), store.NewCapacityRepo(db), peers, exec)
specBytes, _ := json.Marshal(map[string]any{
"name": spec.Name,
"command": "/bin/true", // placeholder; full HCL dispatch lands in a later phase
})
jobID, nodeID, err := dispatcher.Submit(ctx, runTarget, specBytes, runIDKey)
if err != nil {
if jsonOutput {
_ = printJSON(map[string]any{"status": "failed", "error": err.Error()})
}
return err
}
if jsonOutput {
return printJSON(map[string]any{"id": jobID, "node_id": nodeID, "status": "dispatched"})
}
fmt.Fprintf(cmd.OutOrStdout(), "✓ Job dispatched: %s to %s\n", jobID, nodeID)
return nil
}
res, nodesByHost, err := dispatchDecision(ctx, spec, runTarget)
if err != nil {
logDispatch(nil, err)
if jsonOutput {
_ = printJSON(map[string]any{"status": "failed", "error": err.Error()})
}
return err
}
switch res.mode {
case "remote":
// Scheduler selected a node (or --target pinned one): render
// the systemd unit / PVE container, verify it, and SSH-push.
// C-44: a push failure is an error (no local fallback).
unitPaths, derr := deployRemote(ctx, spec, res, nodesByHost)
logDispatch(res, derr)
if derr != nil {
if jsonOutput {
_ = printJSON(map[string]any{"status": "failed", "node": res.node, "error": derr.Error()})
}
return derr
}
res.unitPaths = unitPaths
// REQ-166 / Phase C2: insert a Job DB record so `job list`
// and `job stop` can find the remotely-deployed job.
if dbErr := insertRemoteJob(spec, res.node); dbErr != nil {
// Non-fatal: the job is deployed, just not visible to list.
logDispatch(res, dbErr)
}
cacheInvalidate(cacheJobClass)
if jsonOutput {
return printJSON(map[string]any{
"status": "deployed",
"node": res.node,
"alloc_id": res.allocID,
"units": unitPaths,
})
}
fmt.Fprintf(cmd.OutOrStdout(), "✓ Job deployed to %s: %s (%s)\n", res.node, spec.Name, strings.Join(unitPaths, ", "))
return nil
case "local":
// Local exec fallback (dev mode: no remote nodes registered).
exec, closer, err := jobExecutor()
if err != nil {
return err
}
defer closer()
job := &model.Job{
ID: uuid.NewString(),
Name: spec.Name,
Spec: args[0],
Status: model.JobStatusPending,
}
runErr := exec.Run(ctx, job, workloadToTaskSpecs(spec))
logDispatch(res, runErr)
// REQ-156 / P07 T5: invalidate the jobs cache so the next
// `orca job list` reflects the just-run (or just-failed)
// job instead of a stale cached list.
cacheInvalidate(cacheJobClass)
if runErr != nil {
if jsonOutput {
_ = printJSON(map[string]any{"id": job.ID, "status": "failed", "error": runErr.Error()})
return runErr
}
fmt.Fprintf(cmd.ErrOrStderr(), "✗ Job %s failed: %v\n", job.ID, runErr)
return runErr
}
if jsonOutput {
return printJSON(map[string]any{"id": job.ID, "name": job.Name, "status": "complete"})
}
fmt.Fprintf(cmd.OutOrStdout(), "✓ Job complete: %s (%s)\n", job.ID, job.Name)
return nil
}
return fmt.Errorf("job run: unknown dispatch mode %q", res.mode)
},
}
var jobListCmd = &cobra.Command{
Use: "list",
Short: "List all jobs",
Long: "Display all jobs and their status. Use --watch to stream updates until Ctrl-C.",
RunE: func(cmd *cobra.Command, args []string) error {
if jobWatch {
return watchJobs(cmd)
}
// Cache (R-008): read path only; --watch bypasses.
var cachedJobs []*model.Job
if cacheGetList(cacheJobClass, cacheListKey, &cachedJobs) {
return renderJobs(cmd, cachedJobs)
}
ctx, cancel := context.WithTimeout(cmd.Context(), 5*time.Second)
defer cancel()
db, closer, err := openDB()
if err != nil {
return err
}
defer closer()
jobs, err := store.NewJobRepo(db).List(ctx)
if err != nil {
return err
}
cachePutList(cacheJobClass, cacheListKey, jobs, cacheJobTTL)
return renderJobs(cmd, jobs)
},
}
// renderJobs prints the job list in either JSON or table form.
func renderJobs(cmd *cobra.Command, jobs []*model.Job) error {
if jsonOutput {
return printJSON(jobs)
}
if len(jobs) == 0 {
fmt.Fprintln(cmd.OutOrStdout(), "No jobs. Use 'orca job run <spec.md>' to submit one.")
return nil
}
fmt.Fprintf(cmd.OutOrStdout(), "%-10s %-20s %-12s %-20s %-5s\n", "ID", "NAME", "STATUS", "NODE", "EXIT")
for _, j := range jobs {
shortID := j.ID
if len(shortID) > 8 {
shortID = shortID[:8]
}
exit := "-"
if j.Status == model.JobStatusComplete || j.Status == model.JobStatusFailed || j.Status == model.JobStatusStopped {
exit = fmt.Sprintf("%d", j.ExitCode)
}
fmt.Fprintf(cmd.OutOrStdout(), "%-10s %-20s %-12s %-20s %-5s\n", shortID, j.Name, j.Status, j.Node, exit)
}
return nil
}
func watchJobs(cmd *cobra.Command) error {
ctx, cancel := signal.NotifyContext(cmd.Context(), os.Interrupt, syscall.SIGTERM)
defer cancel()
return watchJobsCtx(cmd, ctx)
}
func watchJobsCtx(cmd *cobra.Command, ctx context.Context) error {
db, closer, err := openDB()
if err != nil {
return err
}
defer closer()
out := cmd.OutOrStdout()
if jsonOutput {
seen := make(map[string]string)
for snapshot := range store.NewJobRepo(db).Watch(ctx) {
current := make(map[string]bool, len(snapshot))
for _, j := range snapshot {
current[j.ID] = true
compact, _ := json.Marshal(j)
key := string(compact)
if prev, ok := seen[j.ID]; !ok || prev != key {
event := "init"
if ok {
event = "update"
}
line, _ := json.Marshal(map[string]any{"event": event, "job": j})
fmt.Fprintln(out, string(line))
seen[j.ID] = key
}
}
for id := range seen {
if !current[id] {
line, _ := json.Marshal(map[string]any{"event": "delete", "id": id})
fmt.Fprintln(out, string(line))
delete(seen, id)
}
}
}
return nil
}
prevTable := ""
for snapshot := range store.NewJobRepo(db).Watch(ctx) {
table := renderJobTable(snapshot)
if table != prevTable {
fmt.Fprint(out, "\033[2J\033[H")
fmt.Fprint(out, table)
prevTable = table
}
}
return nil
}
func renderJobTable(jobs []*model.Job) string {
if len(jobs) == 0 {
return "No jobs.\n"
}
out := fmt.Sprintf("%-10s %-20s %-12s %-20s %-5s\n", "ID", "NAME", "STATUS", "NODE", "EXIT")
for _, j := range jobs {
shortID := j.ID
if len(shortID) > 8 {
shortID = shortID[:8]
}
exit := "-"
if j.Status == model.JobStatusComplete || j.Status == model.JobStatusFailed || j.Status == model.JobStatusStopped {
exit = fmt.Sprintf("%d", j.ExitCode)
}
out += fmt.Sprintf("%-10s %-20s %-12s %-20s %-5s\n", shortID, j.Name, j.Status, j.Node, exit)
}
return out
}
// jobStopTransport is the SSH command-execution seam used by
// `orca job stop`. *sshpush.Transport satisfies it via Exec; tests
// inject a mock (same pattern as driftTransport / drainExecer).
type jobStopTransport interface {
Exec(ctx context.Context, peer string, cmd string) ([]byte, error)
}
// jobStopTransportOverride is the package-level test seam for the
// SSH transport used by `orca job stop`. When non-nil it replaces the
// production transport; tests set it and restore nil in cleanup.
var jobStopTransportOverride jobStopTransport
// jobStopTimeout is the SSH command timeout for `orca job stop`.
var jobStopTimeout time.Duration
// jobStopPeer is the optional --peer override for `orca job stop`.
// When empty, the node is looked up from the alloc_history table
// (latest entry for the job id). When set, the SSH stop targets that
// peer directly.
var jobStopPeer string
func jobStopTransportFromCtx() (jobStopTransport, error) {
if jobStopTransportOverride != nil {
return jobStopTransportOverride, nil
}
keyPath := certpaths.SSHKeyPath()
khPath := certpaths.KnownHostsPath()
return sshpush.NewTransport(keyPath, khPath), nil
}
// nodeForJob looks up the node that ran (or is running) a job by
// searching the alloc_history table for the latest entry for the
// given job id. Returns nil if no history entry exists (the job may
// have been run locally or pre-dates alloc_history).
func nodeForJob(ctx context.Context, db *sql.DB, jobID string) (*model.Node, error) {
hist := store.NewAllocHistoryRepo(db)
if err := hist.EnsureSchema(ctx); err != nil {
return nil, fmt.Errorf("alloc history schema: %w", err)
}
entries, err := hist.List(ctx, store.HistoryFilter{JobID: jobID})
if err != nil {
return nil, fmt.Errorf("alloc history list: %w", err)
}
if len(entries) == 0 {
return nil, nil
}
// Pick the latest entry (List returns ASC; take the last).
latest := entries[len(entries)-1]
if latest.NodeID == "" {
return nil, nil
}
nodeRepo := store.NewNodeRepo(db)
n, err := nodeRepo.Get(ctx, latest.NodeID)
if err != nil {
if errors.Is(err, store.ErrNotFound) {
return nil, nil
}
return nil, fmt.Errorf("lookup node %s: %w", latest.NodeID, err)
}
return n, nil
}
var jobStopCmd = &cobra.Command{
Use: "stop [job-id]",
Short: "Stop a running job",
Long: `Stop a running job by sending 'systemctl stop orca-alloc-<name>-*'
to the node running the allocation via SSH, then mark the job as
stopped in the DB (REQ-158, P09 T1).
If --peer is not given, the node is looked up from the allocation
history. If no node is found, the DB status is updated anyway (soft
stop fallback for local-run jobs).`,
Args: cobra.MaximumNArgs(1),
RunE: func(cmd *cobra.Command, args []string) error {
id := stopID
if id == "" && len(args) > 0 {
id = args[0]
}
if id == "" {
return fmt.Errorf("job id required (--id or argument)")
}
db, closer, err := openDB()
if err != nil {
return err
}
defer closer()
ctx, cancel := context.WithTimeout(cmd.Context(), 30*time.Second)
defer cancel()
repo := store.NewJobRepo(db)
job, err := repo.Get(ctx, id)
if err != nil {
if errors.Is(err, store.ErrNotFound) {
return fmt.Errorf("job not found: %s", id)
}
return err
}
// Determine the peer to SSH to. --peer takes precedence;
// otherwise look up the node from alloc_history.
peer := jobStopPeer
var node *model.Node
if peer == "" {
node, err = nodeForJob(ctx, db, id)
if err != nil {
return fmt.Errorf("lookup node for job %s: %w", id, err)
}
if node != nil {
peer = peerAddrForNode(node)
}
}
// Validate the job name before interpolation into the shell
// command (same injection guard as logs --job / stopAlloc).
jobName := job.Name
if !validSafeName(jobName) {
return fmt.Errorf("job stop: invalid job name %q (allowed: A-Z a-z 0-9 _ -)", jobName)
}
sshRan := false
if peer != "" {
transport, terr := jobStopTransportFromCtx()
if terr != nil {
return fmt.Errorf("job stop: ssh transport: %w", terr)
}
stopCtx, stopCancel := sshCmdCtx(ctx, jobStopTimeout)
defer stopCancel()
// Match the drift.go job restart unit pattern: orca-alloc-<name>.
// Use a glob (orca-alloc-<name>-*) to stop all task units in
// a multi-task allocation group.
unitPattern := fmt.Sprintf("orca-alloc-%s-*", jobName)
stopCmd := fmt.Sprintf("systemctl stop %s", shellQuote(unitPattern))
out, sErr := transport.Exec(stopCtx, peer, stopCmd)
if sErr != nil {
// Non-fatal: the unit may not be running (already
// stopped) or SSH may fail. We still update the DB
// status so the operator's intent is recorded.
fmt.Fprintf(cmd.ErrOrStderr(), "⚠ job stop: SSH systemctl stop failed on %s: %v (output: %s)\n", peer, sErr, strings.TrimSpace(string(out)))
} else {
sshRan = true
}
}
if err := repo.UpdateStatus(ctx, id, model.JobStatusStopped, 130); err != nil {
return err
}
// REQ-156 / P07 T5: invalidate the jobs cache so the next
// `orca job list` reflects the just-stopped job.
cacheInvalidate(cacheJobClass)
if jsonOutput {
result := map[string]any{"id": id, "status": "stopped", "previous_status": job.Status}
if peer != "" {
result["peer"] = peer
result["ssh_stop"] = sshRan
}
return printJSON(result)
}
if peer != "" && sshRan {
fmt.Fprintf(cmd.OutOrStdout(), "✓ Job stopped: %s (systemctl stop on %s)\n", id, peer)
} else if peer != "" {
fmt.Fprintf(cmd.OutOrStdout(), "✓ Job stopped: %s (DB only; SSH stop failed — see stderr)\n", id)
} else {
fmt.Fprintf(cmd.OutOrStdout(), "✓ Job stopped: %s (DB only; no node found)\n", id)
}
return nil
},
}
var jobLogsCmd = &cobra.Command{
Use: "logs [job-id]",
Short: "Show task output for a job",
Long: "Display captured stdout/stderr for all tasks in a job.",
Args: cobra.MaximumNArgs(1),
RunE: func(cmd *cobra.Command, args []string) error {
id := stopID
if id == "" && len(args) > 0 {
id = args[0]
}
if id == "" {
return fmt.Errorf("job id required (--id or argument)")
}
ctx, cancel := context.WithTimeout(cmd.Context(), 5*time.Second)
defer cancel()
db, closer, err := openDB()
if err != nil {
return err
}
defer closer()
taskRepo := store.NewTaskRepo(db)
tasks, err := taskRepo.ListByJob(ctx, id)
if err != nil {
return err
}
if jsonOutput {
return printJSON(tasks)
}
if len(tasks) == 0 {
fmt.Fprintln(cmd.OutOrStdout(), "No tasks for this job.")
return nil
}
for i, t := range tasks {
fmt.Fprintf(cmd.OutOrStdout(), "--- task[%d] %s (%s) exit=%d ---\n", i, t.Command, t.Status, t.ExitCode)
if t.Stdout != "" {
fmt.Fprintln(cmd.OutOrStdout(), t.Stdout)
}
if t.Stderr != "" {
fmt.Fprintln(cmd.OutOrStderr(), t.Stderr)
}
}
return nil
},
}
func init() {
jobStopCmd.Flags().StringVar(&stopID, "id", "", "job id")
jobStopCmd.Flags().StringVar(&jobStopPeer, "peer", "", "peer address (host:port) running the allocation (auto-detected from alloc history if empty)")
jobStopCmd.Flags().DurationVar(&jobStopTimeout, "timeout", sshCmdDefaultTimeout, "SSH command timeout")
jobLogsCmd.Flags().StringVar(&stopID, "id", "", "job id")
jobRunCmd.Flags().StringVar(&runTarget, "target", "", "pin job to a specific node id (overrides bin-packing)")
jobRunCmd.Flags().StringVar(&runIDKey, "idempotency-key", "", "X-Orca-Idempotency-Key for cross-node dispatch dedupe")
jobListCmd.Flags().BoolVar(&jobWatch, "watch", false, "stream jobs until Ctrl-C (table refresh or --json per-event)")
jobCmd.AddCommand(jobRunCmd)
jobCmd.AddCommand(jobListCmd)
jobCmd.AddCommand(jobStopCmd)
jobCmd.AddCommand(jobLogsCmd)
rootCmd.AddCommand(jobCmd)
}
func toTaskSpecs(in []jobspec.TaskSpec) []engine.TaskSpec {
out := make([]engine.TaskSpec, len(in))
for i, t := range in {
out[i] = engine.TaskSpec{
Name: t.Name,
Command: t.Command,
Args: t.Args,
Env: t.Env,
}
}
return out
}
// workloadToTaskSpecs converts a *WorkloadSpec into the engine.TaskSpec
// slice consumed by the executor. For the HCL adapter path the runtime
// block carries the legacy task[0].Command; for the Markdown path the
// runtime block is the canonical runtime abstraction (P07 will expand
// this). When Runtime is nil we emit a single no-op task to preserve
// the legacy "at least one task" invariant.
//
// The runtime command string is split into binary + args via
// splitCommand so that exec.Command receives the binary path and the
// args as separate elements. Without this split, a command like
// "/usr/bin/httpd -f /etc/orca/web-app/httpd.conf" is treated as a
// single file path and fork/exec fails with "no such file or directory".
func workloadToTaskSpecs(spec *jobspec.WorkloadSpec) []engine.TaskSpec {
if spec == nil {
return nil
}
if spec.Runtime == nil {
return []engine.TaskSpec{{Name: spec.Name, Command: "/bin/true"}}
}
bin, args := splitCommand(spec.Runtime.Command)
return []engine.TaskSpec{{
Name: spec.Name,
Command: bin,
Args: args,
}}
}
// splitCommand splits a command string into binary + args using
// strings.Fields (handles multiple spaces/tabs). If the string is empty
// or all-whitespace, returns ("/bin/true", nil) so the executor still
// has a valid binary to run.
func splitCommand(s string) (string, []string) {
parts := strings.Fields(s)
if len(parts) == 0 {
return "/bin/true", nil
}
return parts[0], parts[1:]
}
// insertRemoteJob inserts a model.Job row for a remotely-deployed job
// (REQ-166, Phase C2). Without this, `job list` shows nothing for remote
// deployments and `job stop` can't find the node.
func insertRemoteJob(spec *jobspec.WorkloadSpec, node string) error {
db, closer, err := openDB()
if err != nil {
return err
}
defer closer()
repo := store.NewJobRepo(db)
return repo.Insert(context.Background(), &model.Job{
ID: uuid.NewString(),
Name: spec.Name,
Spec: "",
Status: model.JobStatusRunning,
CreatedAt: time.Now().UTC(),
Node: node,
})
}