From 04902ff24a0ef60dc291d0875c39377db9c94d81 Mon Sep 17 00:00:00 2001 From: jochen Date: Mon, 26 Oct 2026 13:37:00 +0100 Subject: [PATCH] Build seats yield: idle priority and a per-node processor cap on their runtimes, threads per step within it (issue 1806) The operator's desktop slows while it builds: the Go build runtime took 1650% of shanks' 24 processors with nothing holding it back. A build seat's holder now runs its runtime at idle priority under its node's cpu-cap share (declaration Yields, composed from Seat.Yields and the module's cpu-cap setting), and buildcpu/buildstep give each step its part of that cap in threads. --- go/buildcpu/buildcpu.go | 258 +++++++++++++++++++++++++++++++++ go/buildcpu/buildcpu_test.go | 92 ++++++++++++ go/buildstep/capacity_test.go | 16 ++ go/buildstep/cpu.go | 40 +++++ go/catalogue/runtime.go | 8 + go/catalogue/runtime_test.go | 28 ++++ go/catalogue/seats.go | 11 +- go/catalogue/yields.go | 41 ++++++ go/declaration/declaration.go | 43 ++++++ go/declaration/process_test.go | 19 +++ 10 files changed, 553 insertions(+), 3 deletions(-) create mode 100644 go/buildcpu/buildcpu.go create mode 100644 go/buildcpu/buildcpu_test.go create mode 100644 go/buildstep/cpu.go create mode 100644 go/catalogue/yields.go diff --git a/go/buildcpu/buildcpu.go b/go/buildcpu/buildcpu.go new file mode 100644 index 0000000..4df094e --- /dev/null +++ b/go/buildcpu/buildcpu.go @@ -0,0 +1,258 @@ +// Package buildcpu is how much of a build node's processors its builds may take, and how they yield to everything else +// there (novox/hq issue 1806; the operator, 2026-10-26: "limit CPU usage on our build nodes so the total cpu usage never +// impacts the normal workings of our nodes"). +// +// Three parts, each with one owner: +// +// - **Priority.** A build seat's holder (catalogue.Seat.Yields) runs at the lowest CPU and IO priority there is: the +// composer gives its runtime process declaration.Yields, and the node-engine writes it into the unit as +// CPUWeight=idle, IOWeight=1, Nice=19 and the best-effort IO class at its lowest level. Interactive and service +// work always comes first; builds take the cycles nothing else wants. +// - **The cap.** The build modules' setting `cpu-cap` (Setting) gives each node the share of its processors its builds +// may take at most, as `=` pairs. The composer turns this node's share into Yields.CPUShare, and the +// node-engine into CPUQuota for the machine's processor count: a hard bound, held by the kernel. +// - **Threads.** Inside that bound a step starts no more threads than its part of it (Threads): GOMAXPROCS, `go +// test -p`, the Node test runner's concurrency. A step with more runnable threads than its quota only waits in the +// kernel's throttle, and its timeouts with it. +package buildcpu + +import ( + "fmt" + "os" + "runtime" + "sort" + "strconv" + "strings" +) + +// Setting is the build modules' setting of each node's processor cap: `=` pairs separated by commas, +// such as `shanks=50,ace=75`; a percent sign after the number is allowed. Each build module declares it with Defaults +// as its default. +const Setting = "cpu-cap" + +// Defaults is the setting's default: the operator's two desktops (shanks, g14) keep half their processors for +// themselves at all times, and ace, a server, builds on three quarters. A node it does not name has DefaultPercent. +const Defaults = "shanks=50,g14=50,ace=75" + +// DefaultPercent is the cap of a node the setting does not name. +const DefaultPercent = 50 + +// Parse reads the setting's pairs. A percent outside 1 to 100 is refused: 0 would stop every build without pausing the +// seat (the capacity setting says that, with 0 asks), and more than 100 is no cap. +func Parse(s string) (map[string]int, error) { + out := map[string]int{} + for _, pair := range strings.Split(s, ",") { + pair = strings.TrimSpace(pair) + if pair == "" { + continue + } + node, value, ok := strings.Cut(pair, "=") + node = strings.TrimSpace(node) + value = strings.TrimSuffix(strings.TrimSpace(value), "%") + n, err := strconv.Atoi(strings.TrimSpace(value)) + if !ok || node == "" || err != nil || n < 1 || n > 100 { + return nil, fmt.Errorf("the processor cap %q is not =", pair) + } + if _, dup := out[node]; dup { + return nil, fmt.Errorf("the processor cap names %s twice", node) + } + out[node] = n + } + return out, nil +} + +// Of is node's cap in percent of its processors: DefaultPercent for a node the setting does not name. +func Of(setting, node string) (int, error) { + caps, err := Parse(setting) + if err != nil { + return 0, err + } + if n, ok := caps[node]; ok { + return n, nil + } + return DefaultPercent, nil +} + +// Format is a cap map in the setting's own words, sorted by node. +func Format(caps map[string]int) string { + nodes := make([]string, 0, len(caps)) + for n := range caps { + nodes = append(nodes, n) + } + sort.Strings(nodes) + parts := make([]string, 0, len(nodes)) + for _, n := range nodes { + parts = append(parts, fmt.Sprintf("%s=%d", n, caps[n])) + } + return strings.Join(parts, ",") +} + +// Quota is the unit's CPUQuota for a share of a machine with cores processors: systemd counts a quota in percent of +// one processor, so half of 24 is 1200%. At least 100%, one processor, so a cap on a small machine never stops a step. +func Quota(percent, cores int) int { + q := percent * cores + if q < 100 { + q = 100 + } + return q +} + +// Threads is how many threads one step may run at once: its part of the cap when slots steps run beside it, at least +// one. Rounded down, so the steps together stay inside the quota. +func Threads(percent, cores, slots int) int { + if slots < 1 { + slots = 1 + } + if percent < 1 || percent > 100 { + percent = 100 + } + t := cores * percent / 100 / slots + if t < 1 { + t = 1 + } + return t +} + +// Limits is what one step is held to on this node, as a status tool says it. +type Limits struct { + // Percent is the node's cap in percent of its processors, and Setting the words it came from. + Percent int `json:"cap-percent"` + Setting string `json:"cap-setting"` + // Cores is the machine's processor count. + Cores int `json:"cores"` + // Slots is the node's build capacity: steps at once. + Slots int `json:"slots"` + // Threads is what each step is given (Threads). + Threads int `json:"threads-per-step"` +} + +// LimitsFor is the limits of a step on node, from the cap setting and the node's capacity. A setting that cannot be read +// gives the default cap and says why in the error, so a step still runs with a bound rather than none. +func LimitsFor(setting, node string, slots int) (Limits, error) { + l := Limits{Setting: setting, Cores: runtime.NumCPU(), Slots: slots} + p, err := Of(setting, node) + if err != nil { + p = DefaultPercent + } + l.Percent = p + l.Threads = Threads(p, l.Cores, slots) + return l, err +} + +// Env is the environment a step's commands are given so they start no more threads than they may: GOMAXPROCS for every +// Go program (the compiler, a test binary, esbuild, buildah), UV_THREADPOOL_SIZE for Node's worker pool. Returned as +// KEY=value words to append to a command's environment. +func Env(threads int) []string { + if threads < 1 { + threads = 1 + } + t := strconv.Itoa(threads) + return []string{"GOMAXPROCS=" + t, "UV_THREADPOOL_SIZE=" + t} +} + +// WithEnv is env with Env(threads) in place of any GOMAXPROCS or UV_THREADPOOL_SIZE it carried. +func WithEnv(env []string, threads int) []string { + out := make([]string, 0, len(env)+2) + for _, kv := range env { + if strings.HasPrefix(kv, "GOMAXPROCS=") || strings.HasPrefix(kv, "UV_THREADPOOL_SIZE=") { + continue + } + out = append(out, kv) + } + return append(out, Env(threads)...) +} + +// InForce is what the kernel holds this process to, read rather than assumed (evidence over assertion): its nice +// level, its cgroup, the cgroup's CPU weight, idle mark and quota. A field the machine does not answer is left empty +// and Unread says which. +type InForce struct { + Nice int `json:"nice"` + Cgroup string `json:"cgroup,omitempty"` + CPUIdle string `json:"cpu-idle,omitempty"` + CPUWeight string `json:"cpu-weight,omitempty"` + // CPUMax is the cgroup's cpu.max: " ", or "max" with no quota. + CPUMax string `json:"cpu-max,omitempty"` + // Quota is CPUMax as a percent of one processor, 0 with none. + Quota int `json:"quota-percent,omitempty"` + IOWeight string `json:"io-weight,omitempty"` + Unread []string `json:"unread,omitempty"` +} + +// cgroupRoot is where the unified hierarchy is mounted. A variable so a test can point it elsewhere. +var cgroupRoot = "/sys/fs/cgroup" + +// procSelfCgroup is this process's cgroup file. A variable so a test can point it elsewhere. +var procSelfCgroup = "/proc/self/cgroup" + +// ReadInForce reads what the kernel holds this process to. +func ReadInForce() InForce { + var f InForce + f.Nice = nice() + raw, err := os.ReadFile(procSelfCgroup) + if err != nil { + f.Unread = append(f.Unread, "cgroup: "+err.Error()) + return f + } + for _, line := range strings.Split(strings.TrimSpace(string(raw)), "\n") { + if strings.HasPrefix(line, "0::") { + f.Cgroup = strings.TrimPrefix(line, "0::") + } + } + if f.Cgroup == "" { + f.Unread = append(f.Unread, "cgroup: not on the unified hierarchy") + return f + } + dir := cgroupRoot + f.Cgroup + read := func(name string) string { + b, err := os.ReadFile(dir + "/" + name) + if err != nil { + f.Unread = append(f.Unread, name+": "+err.Error()) + return "" + } + return strings.TrimSpace(string(b)) + } + f.CPUIdle = read("cpu.idle") + f.CPUWeight = read("cpu.weight") + f.CPUMax = read("cpu.max") + f.IOWeight = read("io.weight") + f.Quota = quotaOf(f.CPUMax) + return f +} + +// quotaOf is a cpu.max value as a percent of one processor: "1200000 100000" is 1200; "max ..." or nothing is 0. +func quotaOf(cpuMax string) int { + fields := strings.Fields(cpuMax) + if len(fields) != 2 || fields[0] == "max" { + return 0 + } + q, err1 := strconv.Atoi(fields[0]) + p, err2 := strconv.Atoi(fields[1]) + if err1 != nil || err2 != nil || p <= 0 { + return 0 + } + return q * 100 / p +} + +// procSelfStat is this process's stat file. A variable so a test can point it elsewhere. +var procSelfStat = "/proc/self/stat" + +// nice is this process's nice level from its stat file (the 19th field, counted after the command's closing +// parenthesis, which may itself hold spaces); 0 when it cannot be read. +func nice() int { + raw, err := os.ReadFile(procSelfStat) + if err != nil { + return 0 + } + s := string(raw) + i := strings.LastIndex(s, ")") + if i < 0 { + return 0 + } + fields := strings.Fields(s[i+1:]) + // After the command: state is field 3, so nice (field 19) is index 16. + if len(fields) < 17 { + return 0 + } + n, _ := strconv.Atoi(fields[16]) + return n +} diff --git a/go/buildcpu/buildcpu_test.go b/go/buildcpu/buildcpu_test.go new file mode 100644 index 0000000..d1d0549 --- /dev/null +++ b/go/buildcpu/buildcpu_test.go @@ -0,0 +1,92 @@ +package buildcpu + +import ( + "os" + "path/filepath" + "reflect" + "testing" +) + +// The cap's words, a node's share, the unit quota and the threads a step is given, in one table (issue 1806). +func TestCapQuotaAndThreads(t *testing.T) { + for _, c := range []struct { + name, setting, node string + cores, slots int + percent, quota int + threads int + refused bool + }{ + {name: "default for shanks", setting: Defaults, node: "shanks", cores: 24, slots: 3, percent: 50, quota: 1200, threads: 4}, + {name: "default for ace", setting: Defaults, node: "ace", cores: 24, slots: 2, percent: 75, quota: 1800, threads: 9}, + {name: "default for g14", setting: Defaults, node: "g14", cores: 16, slots: 1, percent: 50, quota: 800, threads: 8}, + {name: "a node not named", setting: Defaults, node: "novox", cores: 8, slots: 1, percent: DefaultPercent, quota: 400, threads: 4}, + {name: "percent sign and spaces", setting: " shanks = 25% ", node: "shanks", cores: 24, slots: 3, percent: 25, quota: 600, threads: 2}, + {name: "one processor at least", setting: "tiny=10", node: "tiny", cores: 2, slots: 4, percent: 10, quota: 100, threads: 1}, + {name: "zero slots counts as one", setting: "x=100", node: "x", cores: 4, slots: 0, percent: 100, quota: 400, threads: 4}, + {name: "zero refused", setting: "shanks=0", node: "shanks", refused: true}, + {name: "over a hundred refused", setting: "shanks=150", node: "shanks", refused: true}, + {name: "no number refused", setting: "shanks", node: "shanks", refused: true}, + {name: "twice refused", setting: "a=1,a=2", node: "a", refused: true}, + } { + t.Run(c.name, func(t *testing.T) { + p, err := Of(c.setting, c.node) + if c.refused { + if err == nil { + t.Fatalf("Of(%q) = %d, want refused", c.setting, p) + } + return + } + if err != nil { + t.Fatal(err) + } + if p != c.percent { + t.Fatalf("percent = %d, want %d", p, c.percent) + } + if q := Quota(p, c.cores); q != c.quota { + t.Errorf("Quota = %d, want %d", q, c.quota) + } + if th := Threads(p, c.cores, c.slots); th != c.threads { + t.Errorf("Threads = %d, want %d", th, c.threads) + } + }) + } + caps, _ := Parse("ace=75,shanks=50%") + if got := Format(caps); got != "ace=75,shanks=50" { + t.Errorf("Format = %q", got) + } + if got := WithEnv([]string{"A=1", "GOMAXPROCS=24"}, 3); !reflect.DeepEqual(got, []string{"A=1", "GOMAXPROCS=3", "UV_THREADPOOL_SIZE=3"}) { + t.Errorf("WithEnv = %q", got) + } +} + +// What the kernel holds a process to is read from its cgroup and stat files, not assumed. +func TestReadInForce(t *testing.T) { + dir := t.TempDir() + cg := filepath.Join(dir, "cg") + unit := filepath.Join(cg, "system.slice", "mesh-build-go-runtime.service") + if err := os.MkdirAll(unit, 0o755); err != nil { + t.Fatal(err) + } + for name, v := range map[string]string{"cpu.idle": "1\n", "cpu.weight": "0\n", "cpu.max": "1200000 100000\n", "io.weight": "default 1\n"} { + if err := os.WriteFile(filepath.Join(unit, name), []byte(v), 0o644); err != nil { + t.Fatal(err) + } + } + self := filepath.Join(dir, "cgroup") + stat := filepath.Join(dir, "stat") + _ = os.WriteFile(self, []byte("0::/system.slice/mesh-build-go-runtime.service\n"), 0o644) + // A command with a space and a parenthesis in it, then nice 19 in field 19. + _ = os.WriteFile(stat, []byte("42 (mesh (build) go) S 1 42 42 0 -1 4194560 1 0 0 0 0 0 0 0 39 19 8 0 1 2 3\n"), 0o644) + cgroupRoot, procSelfCgroup, procSelfStat = cg, self, stat + t.Cleanup(func() { cgroupRoot, procSelfCgroup, procSelfStat = "/sys/fs/cgroup", "/proc/self/cgroup", "/proc/self/stat" }) + + got := ReadInForce() + want := InForce{Nice: 19, Cgroup: "/system.slice/mesh-build-go-runtime.service", CPUIdle: "1", CPUWeight: "0", + CPUMax: "1200000 100000", Quota: 1200, IOWeight: "default 1"} + if !reflect.DeepEqual(got, want) { + t.Fatalf("ReadInForce = %+v\nwant %+v", got, want) + } + if q := quotaOf("max 100000"); q != 0 { + t.Errorf("no quota read as %d", q) + } +} diff --git a/go/buildstep/capacity_test.go b/go/buildstep/capacity_test.go index 5334226..8f1a7ac 100644 --- a/go/buildstep/capacity_test.go +++ b/go/buildstep/capacity_test.go @@ -36,3 +36,19 @@ func TestNodeCapacity(t *testing.T) { t.Errorf("a missing file = %d, want 0", got) } } + +// Issue 1806: a step's limits come from the cap in the same settings file, read at every call; a cap that cannot be +// read still bounds the step, at the default. +func TestNodeLimits(t *testing.T) { + file := filepath.Join(t.TempDir(), "settings.json") + os.WriteFile(file, []byte(`{"capacity":"shanks=3","cpu-cap":"shanks=25"}`), 0o600) + if l := NodeLimits(file, "shanks", 3, nil); l.Percent != 25 || l.Slots != 3 || l.Threads < 1 { + t.Errorf("from the file: %+v", l) + } + for _, body := range []string{`{"capacity":"shanks=3"}`, `{"cpu-cap":"shanks=0"}`, `not json`} { + os.WriteFile(file, []byte(body), 0o600) + if l := NodeLimits(file, "shanks", 1, nil); l.Percent != 50 { + t.Errorf("%s gives %d%%, want the default 50", body, l.Percent) + } + } +} diff --git a/go/buildstep/cpu.go b/go/buildstep/cpu.go new file mode 100644 index 0000000..754cd3e --- /dev/null +++ b/go/buildstep/cpu.go @@ -0,0 +1,40 @@ +package buildstep + +import ( + "encoding/json" + "os" + + "git.novox.be/novox/mesh-sdk/go/buildcpu" +) + +// NodeLimits is what one step on node is held to (novox/hq issue 1806): the node's processor cap from the module's +// settings file at path (buildcpu.Setting, its default when the file does not set it), the machine's processors, and +// the node's capacity, so the steps that run at once start no more threads together than the cap gives. Read at every +// call, like the capacity, so a change takes effect at the next step. A file or a setting that cannot be read gives the +// default cap and says why: a step still runs with a bound, never without one. +func NodeLimits(path, node string, slots int, logf func(string, ...any)) buildcpu.Limits { + setting := buildcpu.Defaults + if path != "" { + if raw, err := os.ReadFile(path); err != nil { + sayTo(logf, "the processor cap cannot be read, so %s builds under the default cap: %v", node, err) + } else { + var s map[string]any + if err := json.Unmarshal(raw, &s); err != nil { + sayTo(logf, "the settings file %s is not JSON, so %s builds under the default cap: %v", path, node, err) + } else if v, ok := s[buildcpu.Setting].(string); ok { + setting = v + } + } + } + l, err := buildcpu.LimitsFor(setting, node, slots) + if err != nil { + sayTo(logf, "%v, so %s builds under the default cap of %d%%", err, node, buildcpu.DefaultPercent) + } + return l +} + +func sayTo(logf func(string, ...any), format string, args ...any) { + if logf != nil { + logf(format, args...) + } +} diff --git a/go/catalogue/runtime.go b/go/catalogue/runtime.go index b0139c8..7aa9e36 100644 --- a/go/catalogue/runtime.go +++ b/go/catalogue/runtime.go @@ -329,6 +329,14 @@ func (r Resolution) ownRuntimes(with Rendering) ([]map[string]any, error) { process[HealthField] = m.RuntimeHealth healthInto(process, m, with) } + // A build seat's holder gives way to everything else on its node (novox/hq issue 1806). + yields, err := r.yieldsOf(m, with) + if err != nil { + return nil, err + } + if yields != nil { + process["yields"] = yields + } if err := artifactsInto(process, RuntimeModule, with); err != nil { return nil, err } diff --git a/go/catalogue/runtime_test.go b/go/catalogue/runtime_test.go index 16c02b4..40f6526 100644 --- a/go/catalogue/runtime_test.go +++ b/go/catalogue/runtime_test.go @@ -498,6 +498,34 @@ func TestAModuleOfItsOwnAccountIsServedByARuntimeOfItsOwn(t *testing.T) { if _, told := env[RuntimeOperatorAccount]; told { t.Error("a runtime of a module's own account is told the operator's account") } + if _, yields := own["yields"]; yields { + t.Errorf("telegram holds no build seat and yields: %v", own["yields"]) + } + // A build seat's holder yields: idle priority, under this node's share of its cpu-cap, the default where no layer + // sets one (issue 1806). + builder := telegram + builder.Claims = []Claim{{Name: "node-build-go"}} + for _, c := range []struct { + layers []Layer + share int + }{ + {nil, 50}, + {[]Layer{{From: "anchor", Values: map[string]any{"cpu-cap": "anchor=30,other=90"}}}, 30}, + } { + with.Settings = SettingsBy{"telegram": c.layers} + runtimes, err := Resolution{Node: "anchor", Modules: []Manifest{builder, goRuntime}}.ownRuntimes(with) + if err != nil { + t.Fatal(err) + } + if got := fmt.Sprint(runtimes[0]["yields"]); got != fmt.Sprintf("map[cpu-share:%d priority:idle]", c.share) { + t.Errorf("a build seat's holder under %v yields %s", c.layers, got) + } + } + with.Settings = SettingsBy{"telegram": {{From: "anchor", Values: map[string]any{"cpu-cap": "anchor=0"}}}} + if _, err := (Resolution{Node: "anchor", Modules: []Manifest{builder, goRuntime}}).ownRuntimes(with); err == nil { + t.Error("a cap of 0 percent was composed") + } + with.Settings = nil // Without the machine's runtime to run it from, it is refused in words. if _, err := (Resolution{Node: "anchor", Modules: []Manifest{telegram}}).ownRuntimes(with); err == nil { t.Error("a module of its own account composed without a runtime program") diff --git a/go/catalogue/seats.go b/go/catalogue/seats.go index 7337d8a..bc43fe1 100644 --- a/go/catalogue/seats.go +++ b/go/catalogue/seats.go @@ -49,6 +49,11 @@ type Seat struct { // order Accepts lists them (novox/hq ADR 0525 §2, ADR 0536 §4): the build seats'. Compiled, never stored, like // Receives. PulledByKind bool + // Yields says the seat's holder gives way to everything else on its node (novox/hq issue 1806): its runtime runs at + // the lowest CPU and IO priority and under its node's processor cap, the holder's setting `cpu-cap` + // (buildcpu.Setting). The build seats': a build must never slow the operator's desktop. Compiled, never stored, like + // Receives. + Yields bool // Serves carries each verb in full — name, description, schema — because a role's tools are the // mesh's to define and an agent's to call (novox/hq ADR 0132, design 33 §2). Serves []Verb @@ -256,11 +261,11 @@ var defaultSeats = append([]Seat{ // its holder taking nothing, kept on the module's own pause key (issue 1887's design limit 2). They replace // node-build-agent, retired with build-agent (ADR 0524 §5). {Name: "node-build-go", Scope: ScopeNode, Accepts: buildAskKinds(), PulledByKind: true, Serves: buildSeatVerbs("Go"), - Decision: "novox/hq ADR 0525, ADR 0536"}, + Yields: true, Decision: "novox/hq ADR 0525, ADR 0536"}, {Name: "node-build-node", Scope: ScopeNode, Accepts: buildAskKinds(), PulledByKind: true, Serves: buildSeatVerbs("Node"), - Decision: "novox/hq ADR 0525, ADR 0536"}, + Yields: true, Decision: "novox/hq ADR 0525, ADR 0536"}, {Name: "node-build-image", Scope: ScopeNode, Accepts: buildAskKinds(), PulledByKind: true, Serves: buildSeatVerbs("image"), - Decision: "novox/hq ADR 0525, ADR 0536"}, + Yields: true, Decision: "novox/hq ADR 0525, ADR 0536"}, // **The mesh's resolvers** (novox/hq ADR 0194, 0196, 0223): every node's internal domain, and // every node and container asks them and nothing else. Replicated since ADR 0223: held on more // than one machine, each answering the same names from the same roster, and every machine's diff --git a/go/catalogue/yields.go b/go/catalogue/yields.go new file mode 100644 index 0000000..e0d0034 --- /dev/null +++ b/go/catalogue/yields.go @@ -0,0 +1,41 @@ +package catalogue + +import ( + "fmt" + + "git.novox.be/novox/mesh-sdk/go/buildcpu" +) + +// yieldsOf is how the runtime of its own a module is served by gives way on this node (novox/hq issue 1806), or nil +// when it does not: a module holding a seat that yields (Seat.Yields, the build seats) runs at idle priority and under +// this node's share of its setting `cpu-cap` (buildcpu.Setting), its default when no layer sets it. +// +// In the node-engine's words (declaration.Yields), which it writes into the unit: the kernel holds the bound, not the +// module. +func (r Resolution) yieldsOf(m Manifest, with Rendering) (map[string]any, error) { + yields := false + for _, c := range m.Claims { + if s, ok := SeatNamed(c.Name); ok && s.Yields { + yields = true + } + } + if !yields { + return nil, nil + } + setting := buildcpu.Defaults + for _, s := range Effective(m, with.Settings[m.Module]) { + if s.Key == buildcpu.Setting { + v, ok := s.Value.(string) + if !ok { + return nil, fmt.Errorf("%s's %s is %v, not = pairs (novox/hq issue 1806)", + m.Module, buildcpu.Setting, s.Value) + } + setting = v + } + } + share, err := buildcpu.Of(setting, r.Node) + if err != nil { + return nil, fmt.Errorf("%s's %s: %w (novox/hq issue 1806)", m.Module, buildcpu.Setting, err) + } + return map[string]any{"priority": "idle", "cpu-share": share}, nil +} diff --git a/go/declaration/declaration.go b/go/declaration/declaration.go index cec208d..314cd58 100644 --- a/go/declaration/declaration.go +++ b/go/declaration/declaration.go @@ -916,6 +916,48 @@ type Process struct { // Health is how this resource is ready (novox/hq ADR 0240 rule 2, Phase B): one kind and its // timing, judged by the node-engine beside liveness. Absent: judged alive or not, and nothing more. Health *Health `json:"health,omitempty"` + + // Yields says this process gives way to everything else on its machine (novox/hq issue 1806): the + // runtime of a build seat's holder, whose builds must never slow the operator's desktop. Absent: the + // machine's defaults. For a process that stays up. + Yields *Yields `json:"yields,omitempty"` +} + +// YieldIdle is the one priority a process yields at: the lowest CPU and IO priority there is. +const YieldIdle = "idle" + +// Yields is how a process gives way, written into its unit by the node-engine (novox/hq issue 1806). +// +// **Priority** "idle": CPUWeight=idle (its cgroup runs only on cycles nothing else wants), IOWeight=1, +// Nice=19 for every process it starts, and the best-effort IO class at its lowest level (never the idle IO +// class, which a busy disk starves for good). **CPUShare** is the most of the machine's processors it may +// take, in percent: the node-engine writes CPUQuota for the processors the machine has, so the bound is +// the kernel's. 0 is no cap. +type Yields struct { + Priority string `json:"priority,omitempty"` + CPUShare int `json:"cpu-share,omitempty"` +} + +func (y *Yields) problems(where string, staysUp bool) []string { + if y == nil { + return nil + } + var problems []string + if !staysUp { + problems = append(problems, where+": yields is for a process that stays up") + } + switch y.Priority { + case "", YieldIdle: + default: + problems = append(problems, fmt.Sprintf("%s: yields priority %q is not one this host knows: idle", where, y.Priority)) + } + if y.CPUShare < 0 || y.CPUShare > 100 { + problems = append(problems, fmt.Sprintf("%s: yields cpu-share %d is not a percent of the machine, 0 to 100", where, y.CPUShare)) + } + if y.Priority == "" && y.CPUShare == 0 { + problems = append(problems, where+": yields says neither a priority nor a cpu-share") + } + return problems } func (d *Process) Identity() string { return d.ID } @@ -948,6 +990,7 @@ func ProcessNameProblem(name string) string { func (d *Process) validate(where string, _ bool) []string { problems := d.Health.problems(where, false, !d.RunOnce && d.Schedule == "") + problems = append(problems, d.Yields.problems(where, !d.RunOnce && d.Schedule == "")...) if problem := ProcessNameProblem(d.Name); problem != "" { problems = append(problems, where+": "+problem) } diff --git a/go/declaration/process_test.go b/go/declaration/process_test.go index 9a7c7e9..02eaff1 100644 --- a/go/declaration/process_test.go +++ b/go/declaration/process_test.go @@ -92,6 +92,20 @@ func TestTheModesAreExclusive(t *testing.T) { if problems := watching.validate("a process", false); len(problems) == 0 { t.Fatal("a scheduled process was given something to restart on, and it is never running") } + + // A process yields only while it stays up, at a priority the host knows, within the machine (issue 1806). + for name, bad := range map[string]func(*Process){ + "a step that yields": func(p *Process) { p.RunOnce = true; p.Yields = &Yields{Priority: YieldIdle} }, + "an unknown priority": func(p *Process) { p.Yields = &Yields{Priority: "low"} }, + "a share over the whole": func(p *Process) { p.Yields = &Yields{CPUShare: 150} }, + "a yield saying nothing": func(p *Process) { p.Yields = &Yields{} }, + } { + p := aProcess() + bad(p) + if problems := p.validate("a process", false); len(problems) == 0 { + t.Errorf("%s was accepted", name) + } + } } // A cadence that is not a cadence is refused near its author, rather than by a machine at the far @@ -118,6 +132,11 @@ func TestEachModeOnItsOwnIsAccepted(t *testing.T) { if problems := every.validate("a process", false); len(problems) != 0 { t.Fatalf("a scheduled process was refused: %v", problems) } + yielding := aProcess() + yielding.Yields = &Yields{Priority: YieldIdle, CPUShare: 50} + if problems := yielding.validate("a process", false); len(problems) != 0 { + t.Fatalf("a daemon yielding at idle under half the machine was refused: %v", problems) + } } // **The one that cannot be escaped, only refused.**