feat(host): register and run subcommands, README, lint fixes

commit: b70b2119a868352b44b5eef1f3c251a50feddcc5
author: brice · 9d ago · 2026-10-12 18:44 UTC
parent: 5d5c671
signature: GPG VerifiedCD88F9E564E0102646FB6077A1A07C7C7C3E349D · arch-gpg Verified for the committer email

7 files · +190 -18

M README.md +35 -0


@@ -48,6 +48,41 @@ The host sends a `Cancel{reason}` frame (`execstream.Drive` takes a cancel chann
48	48
49	49		Everything a job creates (process, TAP or bridge, nftables rules, files, loop mounts, containers, networks, cgroups) is removed by `Stop`, which also runs with a cancelled context, aborts a `Start` in progress, and by `Reap` at start for what a crash left behind. A VM leaves a state record (`<state dir>/vms/<id>.json`) before its first resource. The network policy and the TAP or bridge are never removed while the VMM or the container may still be running: when a process cannot be killed or a container cannot be removed, `Stop` returns an error wrapping `sandbox.ErrLeaked`, the sandbox's slot stays taken, and `Driver.Reap` (which returns `sandbox.ErrBusy` while a sandbox is running) finishes the job and frees the slot. The host should stop taking jobs on a leaked slot until then. `Sandbox.ID()` names a sandbox's files and Docker objects.
50	50
	51	+	## Running jobs
	52	+
	53	+	```
	54	+	ermite-runner register --token <token with runner:register> # once
	55	+	ermite-runner run # as a service
	56	+	```
	57	+
	58	+	`register` picks the isolation mode like `run` will, registers the runner with
	59	+	ermite (name, labels, version, isolation, architecture) and stores the runner
	60	+	token in `$ERMITE_RUNNER_STATE_DIR/runner.json` (mode 0600; a file readable by
	61	+	others is refused). The register token can also come from
	62	+	`ERMITE_RUNNER_REGISTER_TOKEN`; `--replace` removes a runner of the same name
	63	+	first.
	64	+
	65	+	`run` reaps what a previous run left, then polls `FetchJob` whenever a slot is
	66	+	free (`ERMITE_RUNNER_CAPACITY`), with a `poll_id` that is kept when a response
	67	+	is lost and obeying `retry_after`. For each job it sends the acknowledging
	68	+	heartbeat first, then one every 30 s (a failed one is retried every 5 s), opens
	69	+	`WatchJob` (a cancel request reaches the executor, which interrupts, terminates
	70	+	and kills like the official runner), forwards the log with `AppendLog`
	71	+	(batches of at most 1,000 lines and 64 KiB per step, numbered offsets, retried
	72	+	with the same lines), sends step snapshots with `UpdateJob` and ends with an
	73	+	idempotent `CompleteJob`. When ermite says the job ended (`JobEnded`) or revokes
	74	+	the assignment (`unauthenticated`) the sandbox is destroyed and nothing more is
	75	+	sent. On SIGINT or SIGTERM the running jobs are cancelled with
	76	+	`runner_shutdown`, reported, and the sandboxes removed. The job token stays in
	77	+	the host process; the sandbox receives the workflow and its contexts only.
	78	+
	79	+	\| Variable \| Meaning \|
	80	+	\| --- \| --- \|
	81	+	\| `ERMITE_RUNNER_URL` \| base URL of ermite \|
	82	+	\| `ERMITE_RUNNER_NAME` \| runner name (default: the host name) \|
	83	+	\| `ERMITE_RUNNER_LABELS` \| extra labels, comma separated (`ubuntu-latest,ubuntu-24.04`) \|
	84	+	\| `ERMITE_RUNNER_REGISTER_TOKEN` \| token for `register` \|
	85	+
51	86		## Configuration (sandbox part)
52	87
53	88		\| Variable \| Default \| \|


@@ -48,6 +48,41 @@ The host sends a `Cancel{reason}` frame (`execstream.Drive` takes a cancel chann
48		48
49	Everything a job creates (process, TAP or bridge, nftables rules, files, loop mounts, containers, networks, cgroups) is removed by `Stop`, which also runs with a cancelled context, aborts a `Start` in progress, and by `Reap` at start for what a crash left behind. A VM leaves a state record (`<state dir>/vms/<id>.json`) before its first resource. The network policy and the TAP or bridge are never removed while the VMM or the container may still be running: when a process cannot be killed or a container cannot be removed, `Stop` returns an error wrapping `sandbox.ErrLeaked`, the sandbox's slot stays taken, and `Driver.Reap` (which returns `sandbox.ErrBusy` while a sandbox is running) finishes the job and frees the slot. The host should stop taking jobs on a leaked slot until then. `Sandbox.ID()` names a sandbox's files and Docker objects.	49		Everything a job creates (process, TAP or bridge, nftables rules, files, loop mounts, containers, networks, cgroups) is removed by `Stop`, which also runs with a cancelled context, aborts a `Start` in progress, and by `Reap` at start for what a crash left behind. A VM leaves a state record (`<state dir>/vms/<id>.json`) before its first resource. The network policy and the TAP or bridge are never removed while the VMM or the container may still be running: when a process cannot be killed or a container cannot be removed, `Stop` returns an error wrapping `sandbox.ErrLeaked`, the sandbox's slot stays taken, and `Driver.Reap` (which returns `sandbox.ErrBusy` while a sandbox is running) finishes the job and frees the slot. The host should stop taking jobs on a leaked slot until then. `Sandbox.ID()` names a sandbox's files and Docker objects.
50		50
		51	+	## Running jobs
		52	+
		53	+	```
		54	+	ermite-runner register --token <token with runner:register> # once
		55	+	ermite-runner run # as a service
		56	+	```
		57	+
		58	+	`register` picks the isolation mode like `run` will, registers the runner with
		59	+	ermite (name, labels, version, isolation, architecture) and stores the runner
		60	+	token in `$ERMITE_RUNNER_STATE_DIR/runner.json` (mode 0600; a file readable by
		61	+	others is refused). The register token can also come from
		62	+	`ERMITE_RUNNER_REGISTER_TOKEN`; `--replace` removes a runner of the same name
		63	+	first.
		64	+
		65	+	`run` reaps what a previous run left, then polls `FetchJob` whenever a slot is
		66	+	free (`ERMITE_RUNNER_CAPACITY`), with a `poll_id` that is kept when a response
		67	+	is lost and obeying `retry_after`. For each job it sends the acknowledging
		68	+	heartbeat first, then one every 30 s (a failed one is retried every 5 s), opens
		69	+	`WatchJob` (a cancel request reaches the executor, which interrupts, terminates
		70	+	and kills like the official runner), forwards the log with `AppendLog`
		71	+	(batches of at most 1,000 lines and 64 KiB per step, numbered offsets, retried
		72	+	with the same lines), sends step snapshots with `UpdateJob` and ends with an
		73	+	idempotent `CompleteJob`. When ermite says the job ended (`JobEnded`) or revokes
		74	+	the assignment (`unauthenticated`) the sandbox is destroyed and nothing more is
		75	+	sent. On SIGINT or SIGTERM the running jobs are cancelled with
		76	+	`runner_shutdown`, reported, and the sandboxes removed. The job token stays in
		77	+	the host process; the sandbox receives the workflow and its contexts only.
		78	+
		79	+	\| Variable \| Meaning \|
		80	+	\| --- \| --- \|
		81	+	\| `ERMITE_RUNNER_URL` \| base URL of ermite \|
		82	+	\| `ERMITE_RUNNER_NAME` \| runner name (default: the host name) \|
		83	+	\| `ERMITE_RUNNER_LABELS` \| extra labels, comma separated (`ubuntu-latest,ubuntu-24.04`) \|
		84	+	\| `ERMITE_RUNNER_REGISTER_TOKEN` \| token for `register` \|
		85	+
51	## Configuration (sandbox part)	86		## Configuration (sandbox part)
52		87
53	\| Variable \| Default \| \|	88		\| Variable \| Default \| \|

M go.mod +1 -1


@@ -11,7 +11,6 @@ require (
11	11		github.com/google/uuid v1.6.0
12	12		github.com/moby/moby/api v1.56.1
13	13		github.com/moby/moby/client v0.6.2
14		-	golang.org/x/net v0.58.0
15	14		golang.org/x/sys v0.49.0
16	15		google.golang.org/protobuf v1.36.12
17	16		)

@@ -91,6 +90,7 @@ require (
91	90		golang.org/x/crypto v0.56.0 // indirect
92	91		golang.org/x/exp v0.0.0-20260824195058-e88cd73687aa // indirect
93	92		golang.org/x/mod v0.40.0 // indirect
	93	+	golang.org/x/net v0.58.0 // indirect
94	94		golang.org/x/sync v0.23.0 // indirect
95	95		golang.org/x/term v0.45.0 // indirect
96	96		golang.org/x/text v0.41.0 // indirect

 	github.com/google/uuid v1.6.0
 	github.com/moby/moby/api v1.56.1
 	github.com/moby/moby/client v0.6.2
-	golang.org/x/net v0.58.0
 	golang.org/x/sys v0.49.0
 	google.golang.org/protobuf v1.36.12
 )
 	golang.org/x/crypto v0.56.0 // indirect
 	golang.org/x/exp v0.0.0-20260824195058-e88cd73687aa // indirect
 	golang.org/x/mod v0.40.0 // indirect
+	golang.org/x/net v0.58.0 // indirect
 	golang.org/x/sync v0.23.0 // indirect
 	golang.org/x/term v0.45.0 // indirect
 	golang.org/x/text v0.41.0 // indirect

A host_cmd.go +137 -0

@@ -0,0 +1,137 @@
	1	+	// SPDX-License-Identifier: MIT
	2	+
	3	+	package main
	4	+
	5	+	import (
	6	+	"context"
	7	+	"errors"
	8	+	"flag"
	9	+	"fmt"
	10	+	"net/http"
	11	+	"os"
	12	+	"strings"
	13	+
	14	+	"github.com/bnema/zerowrap"
	15	+
	16	+	"bnema.dev/ermite-api/gen/ermite/ermiteconnect"
	17	+
	18	+	"bnema.dev/ermite-runner/internal/host"
	19	+	"bnema.dev/ermite-runner/internal/sandbox"
	20	+	"bnema.dev/ermite-runner/internal/sandbox/drivers"
	21	+	)
	22	+
	23	+	// loadHostConfig reads the sandbox settings and the host settings together, so
	24	+	// that the state directory and the capacity are the same for both.
	25	+	func loadHostConfig() (sandbox.Config, host.Config, error) {
	26	+	scfg, err := sandbox.LoadConfig(os.Getenv)
	27	+	if err != nil {
	28	+	return scfg, host.Config{}, err
	29	+	}
	30	+	name, _ := os.Hostname()
	31	+	hcfg, err := host.LoadConfig(os.Getenv, name, scfg.StateDir, scfg.Capacity, version)
	32	+	return scfg, hcfg, err
	33	+	}
	34	+
	35	+	func buildDriver(ctx context.Context, scfg sandbox.Config) (sandbox.Driver, error) {
	36	+	return sandbox.Select(ctx, scfg, drivers.Factories(), sandbox.Prober{})
	37	+	}
	38	+
	39	+	// registerCmd is `ermite-runner register [--token T] [--replace]`: it chooses
	40	+	// the isolation mode like `run` will, registers the runner with it, and stores
	41	+	// the runner token (mode 0600) in the state directory. The register token comes
	42	+	// from --token or ERMITE_RUNNER_REGISTER_TOKEN, so that it need not appear in a
	43	+	// process list.
	44	+	func registerCmd(ctx context.Context, args []string) int {
	45	+	fs := flag.NewFlagSet("register", flag.ContinueOnError)
	46	+	token := fs.String("token", "", "token with the runner:register scope (default: $"+host.EnvRegisterToken+")")
	47	+	replace := fs.Bool("replace", false, "replace a runner of the same name")
	48	+	if err := fs.Parse(args); err != nil {
	49	+	return 2
	50	+	}
	51	+	if *token == "" {
	52	+	*token = os.Getenv(host.EnvRegisterToken)
	53	+	}
	54	+	scfg, hcfg, err := loadHostConfig()
	55	+	if err != nil {
	56	+	fmt.Fprintln(os.Stderr, "ermite-runner:", err)
	57	+	return 2
	58	+	}
	59	+	switch {
	60	+	case hcfg.URL == "":
	61	+	fmt.Fprintf(os.Stderr, "ermite-runner: %s is not set\n", host.EnvURL)
	62	+	return 2
	63	+	case *token == "":
	64	+	fmt.Fprintf(os.Stderr, "ermite-runner: give a token with --token or %s\n", host.EnvRegisterToken)
	65	+	return 2
	66	+	}
	67	+	drv, err := buildDriver(ctx, scfg)
	68	+	if err != nil {
	69	+	fmt.Fprintln(os.Stderr, "ermite-runner:", err)
	70	+	return 1
	71	+	}
	72	+	iso := drv.Isolation()
	73	+	_ = drv.Close()
	74	+
	75	+	client := host.NewClient(hcfg.URL, *token, http.DefaultTransport)
	76	+	st, err := host.Register(ctx, hcfg, client, iso, *replace)
	77	+	if err != nil {
	78	+	fmt.Fprintln(os.Stderr, "ermite-runner:", err)
	79	+	return 1
	80	+	}
	81	+	fmt.Printf("registered runner %q (%s) with isolation %s; token stored in %s\n", st.Name, st.RunnerID, iso, host.StateFile(hcfg.StateDir))
	82	+	return 0
	83	+	}
	84	+
	85	+	// runCmd is `ermite-runner run`: poll ermite for jobs and run each one in its
	86	+	// own sandbox until SIGINT or SIGTERM, then cancel the running jobs with
	87	+	// runner_shutdown and report them.
	88	+	func runCmd(ctx context.Context) int {
	89	+	log := zerowrap.FromCtx(ctx)
	90	+	scfg, hcfg, err := loadHostConfig()
	91	+	if err != nil {
	92	+	fmt.Fprintln(os.Stderr, "ermite-runner:", err)
	93	+	return 2
	94	+	}
	95	+	st, err := host.LoadState(hcfg.StateDir)
	96	+	if err != nil {
	97	+	fmt.Fprintln(os.Stderr, "ermite-runner:", err)
	98	+	return 2
	99	+	}
	100	+	url := hcfg.URL
	101	+	if url == "" {
	102	+	url = st.URL
	103	+	}
	104	+	if url == "" {
	105	+	fmt.Fprintf(os.Stderr, "ermite-runner: %s is not set\n", host.EnvURL)
	106	+	return 2
	107	+	}
	108	+	drv, err := buildDriver(ctx, scfg)
	109	+	if err != nil {
	110	+	fmt.Fprintln(os.Stderr, "ermite-runner:", err)
	111	+	return 1
	112	+	}
	113	+	defer func() { _ = drv.Close() }()
	114	+	log.Info().Str("runner", st.Name).Str("isolation", string(drv.Isolation())).Int("capacity", scfg.Capacity).Msg("runner started")
	115	+
	116	+	r := &host.Runner{
	117	+	Cfg: hcfg,
	118	+	Driver: drv,
	119	+	Poll: host.NewClient(url, st.RunnerToken, http.DefaultTransport),
	120	+	Job: func(tok string) ermiteconnect.RunnerServiceClient {
	121	+	return host.NewClient(url, tok, http.DefaultTransport)
	122	+	},
	123	+	Logf: func(format string, args ...any) { log.Info().Msg(strings.TrimSpace(fmt.Sprintf(format, args...))) },
	124	+	JobSpec: nil,
	125	+	}
	126	+	r.Cfg.URL = url
	127	+	if err := r.Run(ctx); err != nil {
	128	+	if errors.Is(err, host.ErrRemoved) {
	129	+	fmt.Fprintln(os.Stderr, "ermite-runner:", err)
	130	+	return 3
	131	+	}
	132	+	fmt.Fprintln(os.Stderr, "ermite-runner:", err)
	133	+	return 1
	134	+	}
	135	+	log.Info().Msg("runner stopped")
	136	+	return 0
	137	+	}

@@ -0,0 +1,137 @@
	1	+	// SPDX-License-Identifier: MIT
	2	+
	3	+	package main
	4	+
	5	+	import (
	6	+	"context"
	7	+	"errors"
	8	+	"flag"
	9	+	"fmt"
	10	+	"net/http"
	11	+	"os"
	12	+	"strings"
	13	+
	14	+	"github.com/bnema/zerowrap"
	15	+
	16	+	"bnema.dev/ermite-api/gen/ermite/ermiteconnect"
	17	+
	18	+	"bnema.dev/ermite-runner/internal/host"
	19	+	"bnema.dev/ermite-runner/internal/sandbox"
	20	+	"bnema.dev/ermite-runner/internal/sandbox/drivers"
	21	+	)
	22	+
	23	+	// loadHostConfig reads the sandbox settings and the host settings together, so
	24	+	// that the state directory and the capacity are the same for both.
	25	+	func loadHostConfig() (sandbox.Config, host.Config, error) {
	26	+	scfg, err := sandbox.LoadConfig(os.Getenv)
	27	+	if err != nil {
	28	+	return scfg, host.Config{}, err
	29	+	}
	30	+	name, _ := os.Hostname()
	31	+	hcfg, err := host.LoadConfig(os.Getenv, name, scfg.StateDir, scfg.Capacity, version)
	32	+	return scfg, hcfg, err
	33	+	}
	34	+
	35	+	func buildDriver(ctx context.Context, scfg sandbox.Config) (sandbox.Driver, error) {
	36	+	return sandbox.Select(ctx, scfg, drivers.Factories(), sandbox.Prober{})
	37	+	}
	38	+
	39	+	// registerCmd is `ermite-runner register [--token T] [--replace]`: it chooses
	40	+	// the isolation mode like `run` will, registers the runner with it, and stores
	41	+	// the runner token (mode 0600) in the state directory. The register token comes
	42	+	// from --token or ERMITE_RUNNER_REGISTER_TOKEN, so that it need not appear in a
	43	+	// process list.
	44	+	func registerCmd(ctx context.Context, args []string) int {
	45	+	fs := flag.NewFlagSet("register", flag.ContinueOnError)
	46	+	token := fs.String("token", "", "token with the runner:register scope (default: $"+host.EnvRegisterToken+")")
	47	+	replace := fs.Bool("replace", false, "replace a runner of the same name")
	48	+	if err := fs.Parse(args); err != nil {
	49	+	return 2
	50	+	}
	51	+	if *token == "" {
	52	+	*token = os.Getenv(host.EnvRegisterToken)
	53	+	}
	54	+	scfg, hcfg, err := loadHostConfig()
	55	+	if err != nil {
	56	+	fmt.Fprintln(os.Stderr, "ermite-runner:", err)
	57	+	return 2
	58	+	}
	59	+	switch {
	60	+	case hcfg.URL == "":
	61	+	fmt.Fprintf(os.Stderr, "ermite-runner: %s is not set\n", host.EnvURL)
	62	+	return 2
	63	+	case *token == "":
	64	+	fmt.Fprintf(os.Stderr, "ermite-runner: give a token with --token or %s\n", host.EnvRegisterToken)
	65	+	return 2
	66	+	}
	67	+	drv, err := buildDriver(ctx, scfg)
	68	+	if err != nil {
	69	+	fmt.Fprintln(os.Stderr, "ermite-runner:", err)
	70	+	return 1
	71	+	}
	72	+	iso := drv.Isolation()
	73	+	_ = drv.Close()
	74	+
	75	+	client := host.NewClient(hcfg.URL, *token, http.DefaultTransport)
	76	+	st, err := host.Register(ctx, hcfg, client, iso, *replace)
	77	+	if err != nil {
	78	+	fmt.Fprintln(os.Stderr, "ermite-runner:", err)
	79	+	return 1
	80	+	}
	81	+	fmt.Printf("registered runner %q (%s) with isolation %s; token stored in %s\n", st.Name, st.RunnerID, iso, host.StateFile(hcfg.StateDir))
	82	+	return 0
	83	+	}
	84	+
	85	+	// runCmd is `ermite-runner run`: poll ermite for jobs and run each one in its
	86	+	// own sandbox until SIGINT or SIGTERM, then cancel the running jobs with
	87	+	// runner_shutdown and report them.
	88	+	func runCmd(ctx context.Context) int {
	89	+	log := zerowrap.FromCtx(ctx)
	90	+	scfg, hcfg, err := loadHostConfig()
	91	+	if err != nil {
	92	+	fmt.Fprintln(os.Stderr, "ermite-runner:", err)
	93	+	return 2
	94	+	}
	95	+	st, err := host.LoadState(hcfg.StateDir)
	96	+	if err != nil {
	97	+	fmt.Fprintln(os.Stderr, "ermite-runner:", err)
	98	+	return 2
	99	+	}
	100	+	url := hcfg.URL
	101	+	if url == "" {
	102	+	url = st.URL
	103	+	}
	104	+	if url == "" {
	105	+	fmt.Fprintf(os.Stderr, "ermite-runner: %s is not set\n", host.EnvURL)
	106	+	return 2
	107	+	}
	108	+	drv, err := buildDriver(ctx, scfg)
	109	+	if err != nil {
	110	+	fmt.Fprintln(os.Stderr, "ermite-runner:", err)
	111	+	return 1
	112	+	}
	113	+	defer func() { _ = drv.Close() }()
	114	+	log.Info().Str("runner", st.Name).Str("isolation", string(drv.Isolation())).Int("capacity", scfg.Capacity).Msg("runner started")
	115	+
	116	+	r := &host.Runner{
	117	+	Cfg: hcfg,
	118	+	Driver: drv,
	119	+	Poll: host.NewClient(url, st.RunnerToken, http.DefaultTransport),
	120	+	Job: func(tok string) ermiteconnect.RunnerServiceClient {
	121	+	return host.NewClient(url, tok, http.DefaultTransport)
	122	+	},
	123	+	Logf: func(format string, args ...any) { log.Info().Msg(strings.TrimSpace(fmt.Sprintf(format, args...))) },
	124	+	JobSpec: nil,
	125	+	}
	126	+	r.Cfg.URL = url
	127	+	if err := r.Run(ctx); err != nil {
	128	+	if errors.Is(err, host.ErrRemoved) {
	129	+	fmt.Fprintln(os.Stderr, "ermite-runner:", err)
	130	+	return 3
	131	+	}
	132	+	fmt.Fprintln(os.Stderr, "ermite-runner:", err)
	133	+	return 1
	134	+	}
	135	+	log.Info().Msg("runner stopped")
	136	+	return 0
	137	+	}

M internal/host/config.go +1 -1


@@ -71,7 +71,7 @@ func LoadConfig(getenv func(string) string, hostname, stateDir string, capacity
71	71		func sanitizeName(s string) string {
72	72		b := []byte(s)
73	73		for i, ch := range b {
74		-	if !(ch >= 'a' && ch <= 'z' \|\| ch >= 'A' && ch <= 'Z' \|\| ch >= '0' && ch <= '9' \|\| ch == '.' \|\| ch == '_' \|\| ch == '-') {
	74	+	if ok := ch >= 'a' && ch <= 'z' \|\| ch >= 'A' && ch <= 'Z' \|\| ch >= '0' && ch <= '9' \|\| ch == '.' \|\| ch == '_' \|\| ch == '-'; !ok {
75	75		b[i] = '-'
76	76		}
77	77		}


@@ -71,7 +71,7 @@ func LoadConfig(getenv func(string) string, hostname, stateDir string, capacity
71		func sanitizeName(s string) string {	71		func sanitizeName(s string) string {
72		b := []byte(s)	72		b := []byte(s)
73		for i, ch := range b {	73		for i, ch := range b {
74	-	if !(ch >= 'a' && ch <= 'z' \|\| ch >= 'A' && ch <= 'Z' \|\| ch >= '0' && ch <= '9' \|\| ch == '.' \|\| ch == '_' \|\| ch == '-') {	74	+	if ok := ch >= 'a' && ch <= 'z' \|\| ch >= 'A' && ch <= 'Z' \|\| ch >= '0' && ch <= '9' \|\| ch == '.' \|\| ch == '_' \|\| ch == '-'; !ok {
75		b[i] = '-'	75		b[i] = '-'
76		}	76		}
77		}	77		}

M internal/host/fake_test.go +5 -9


@@ -14,8 +14,6 @@ import (
14	14		"time"
15	15
16	16		"connectrpc.com/connect"
17		-	"golang.org/x/net/http2"
18		-	"golang.org/x/net/http2/h2c"
19	17
20	18		ermite "bnema.dev/ermite-api/gen/ermite"
21	19		"bnema.dev/ermite-api/gen/ermite/ermiteconnect"

@@ -40,10 +38,9 @@ type fakeServer struct {
40	38		pollCalls int
41	39		retryAfter time.Duration
42	40
43		-	beats []time.Time
44		-	beatHook func(n int) error
45		-	registers []*ermite.RegisterRunnerRequest
46		-	registerTok string
	41	+	beats []time.Time
	42	+	beatHook func(n int) error
	43	+	registers []*ermite.RegisterRunnerRequest
47	44
48	45		lines []string // log text, joined pieces are not merged: one per line
49	46		cont []bool

@@ -283,12 +280,11 @@ func (s fakeServer) WatchJob(ctx context.Context, r connect.Request[ermite.Wat
283	280		}
284	281		}
285	282
286		-	// start serves the handler over h2c on a local port and returns its URL.
	283	+	// start serves the handler over HTTP/1.1 on a local port (Connect streams server messages over it) and returns its URL.
287	284		func (s fakeServer) start(t testing.T) string {
288	285		t.Helper()
289	286		_, h := ermiteconnect.NewRunnerServiceHandler(s)
290		-	srv := httptest.NewUnstartedServer(h2c.NewHandler(h, &http2.Server{}))
291		-	srv.Start()
	287	+	srv := httptest.NewServer(h)
292	288		t.Cleanup(srv.Close)
293	289		return srv.URL
294	290		}

 	"time"
 	"connectrpc.com/connect"
-	"golang.org/x/net/http2"
-	"golang.org/x/net/http2/h2c"
 	ermite "bnema.dev/ermite-api/gen/ermite"
 	"bnema.dev/ermite-api/gen/ermite/ermiteconnect"
 	pollCalls  int
 	retryAfter time.Duration
-	beats       []time.Time
+	beats     []time.Time
-	beatHook    func(n int) error
+	beatHook  func(n int) error
-	registers   []*ermite.RegisterRunnerRequest
+	registers []*ermite.RegisterRunnerRequest
-	registerTok string
 	lines      []string // log text, joined pieces are not merged: one per line
 	cont       []bool
 	}
 }
-// start serves the handler over h2c on a local port and returns its URL.
+// start serves the handler over HTTP/1.1 on a local port (Connect streams server messages over it) and returns its URL.
 func (s *fakeServer) start(t *testing.T) string {
 	t.Helper()
 	_, h := ermiteconnect.NewRunnerServiceHandler(s)
-	srv := httptest.NewUnstartedServer(h2c.NewHandler(h, &http2.Server{}))
+	srv := httptest.NewServer(h)
-	srv.Start()
 	t.Cleanup(srv.Close)
 	return srv.URL
 }

M internal/host/job.go +1 -1


@@ -551,7 +551,7 @@ func validOutputName(n string) bool {
551	551		}
552	552		for i, r := range n {
553	553		letter := r >= 'a' && r <= 'z' \|\| r >= 'A' && r <= 'Z' \|\| r == '_'
554		-	if !(letter \|\| (i > 0 && (r >= '0' && r <= '9' \|\| r == '-'))) {
	554	+	if ok := letter \|\| (i > 0 && (r >= '0' && r <= '9' \|\| r == '-')); !ok {
555	555		return false
556	556		}
557	557		}


@@ -551,7 +551,7 @@ func validOutputName(n string) bool {
551		}	551		}
552		for i, r := range n {	552		for i, r := range n {
553		letter := r >= 'a' && r <= 'z' \|\| r >= 'A' && r <= 'Z' \|\| r == '_'	553		letter := r >= 'a' && r <= 'z' \|\| r >= 'A' && r <= 'Z' \|\| r == '_'
554	-	if !(letter \|\| (i > 0 && (r >= '0' && r <= '9' \|\| r == '-'))) {	554	+	if ok := letter \|\| (i > 0 && (r >= '0' && r <= '9' \|\| r == '-')); !ok {
555		return false	555		return false
556		}	556		}
557		}	557		}

M main.go +10 -6


@@ -4,10 +4,10 @@
4	4		// its own Firecracker micro-VM, or in its own container when the host has no
5	5		// KVM, and executes workflows compatible with GitHub Actions.
6	6		//
7		-	// Subcommands: version; init and exec run inside a sandbox (guest PID 1, and
8		-	// the executor); sandbox-check runs one stub job through the selected isolation
9		-	// mode and prints its output (phase D of the CI-1 plan: the host loop that
10		-	// fetches jobs from ermite comes later).
	7	+	// Subcommands: version; register stores a runner token obtained from ermite;
	8	+	// run fetches jobs from ermite and runs each in a sandbox; init and exec run
	9	+	// inside a sandbox (guest PID 1, and the executor); sandbox-check runs one stub
	10	+	// script through the selected isolation mode and prints its output.
11	11		package main
12	12
13	13		import (

@@ -49,13 +49,17 @@ func run(args []string) int {
49	49		return execCmd(ctx)
50	50		case "sandbox-check":
51	51		return sandboxCheck(ctx, args[1:])
	52	+	case "register":
	53	+	return registerCmd(ctx, args[1:])
	54	+	case "run":
	55	+	return runCmd(ctx)
52	56		}
53	57		usage()
54	58		return 2
55	59		}
56	60
57	61		func usage() {
58		-	fmt.Fprintln(os.Stderr, "usage: ermite-runner version \| sandbox-check [script]\n (inside a sandbox: ermite-runner init \| exec)")
	62	+	fmt.Fprintln(os.Stderr, "usage: ermite-runner version \| register [--token T] [--replace] \| run \| sandbox-check [script]\n (inside a sandbox: ermite-runner init \| exec)")
59	63		}
60	64
61	65		func envOr(key, def string) string {

@@ -65,7 +69,7 @@ func envOr(key, def string) string {
65	69		return def
66	70		}
67	71
68		-	// execCmd is `ermite-runner exec`, the stub executor.
	72	+	// execCmd is `ermite-runner exec`, the executor of a sandbox.
69	73		func execCmd(ctx context.Context) int {
70	74		log := zerowrap.FromCtx(ctx)
71	75		target, err := guest.ParseTarget(os.Getenv("ERMITE_CONNECT"))

 // its own Firecracker micro-VM, or in its own container when the host has no
 // KVM, and executes workflows compatible with GitHub Actions.
 //
-// Subcommands: version; init and exec run inside a sandbox (guest PID 1, and
+// Subcommands: version; register stores a runner token obtained from ermite;
-// the executor); sandbox-check runs one stub job through the selected isolation
+// run fetches jobs from ermite and runs each in a sandbox; init and exec run
-// mode and prints its output (phase D of the CI-1 plan: the host loop that
+// inside a sandbox (guest PID 1, and the executor); sandbox-check runs one stub
-// fetches jobs from ermite comes later).
+// script through the selected isolation mode and prints its output.
 package main
 import (
 		return execCmd(ctx)
 	case "sandbox-check":
 		return sandboxCheck(ctx, args[1:])
+	case "register":
+		return registerCmd(ctx, args[1:])
+	case "run":
+		return runCmd(ctx)
 	}
 	usage()
 	return 2
 }
 func usage() {
-	fmt.Fprintln(os.Stderr, "usage: ermite-runner version | sandbox-check [script]\n       (inside a sandbox: ermite-runner init | exec)")
+	fmt.Fprintln(os.Stderr, "usage: ermite-runner version | register [--token T] [--replace] | run | sandbox-check [script]\n       (inside a sandbox: ermite-runner init | exec)")
 }
 func envOr(key, def string) string {
 	return def
 }
-// execCmd is `ermite-runner exec`, the stub executor.
+// execCmd is `ermite-runner exec`, the executor of a sandbox.
 func execCmd(ctx context.Context) int {
 	log := zerowrap.FromCtx(ctx)
 	target, err := guest.ParseTarget(os.Getenv("ERMITE_CONNECT"))