Read the newest steps first: sort the node's keys by id (start order) before the cut, and say older steps were not read
mesh/delivery merged: source.main moved to novox/mesh-catalog@567736f7795a
mesh/check failed: integration: meshbuildnode/cmd/mesh-build-node.TestAFixtureBundleBuiltTwiceHasOneDigest: no toolchain of this lock is installed und…

The review found a cut read took an arbitrary 1000 keys, so the newest steps could be missing. A step id is
base-36 milliseconds plus random digits, so its string order is its start order; the newest max(limit*4, 100)
keys are read now, 1000 at most. The toolchains read logs when it drops keys past its cap.
This commit is contained in:
2026-10-26 00:43:53 +01:00
parent 8f9c78bcba
commit e901769cb2
6 changed files with 60 additions and 43 deletions
@@ -131,8 +131,9 @@ type StepLine struct {
} }
// RecentSteps are this node's steps, the latest started first, at most limit (ten when limit is not above zero). Only // RecentSteps are this node's steps, the latest started first, at most limit (ten when limit is not above zero). Only
// the node's own keys (<node>.) are read, at most MaxStateReads of them; cut says more were stored than were read, so // the node's own keys (<node>.) are listed and the newest max(limit*4, 100) of them read (at most MaxStateReads): a step's
// the newest may be missing. A step's id says nothing of when it started, so the values are what orders them. // id is base-36 milliseconds and random digits, so its string order is its start order. cut says older steps were not
// read. The values still order the lines, since an ask step's id carries the ask time.
func (m *Module) RecentSteps(limit int) (lines []StepLine, cut bool, err error) { func (m *Module) RecentSteps(limit int) (lines []StepLine, cut bool, err error) {
if limit <= 0 { if limit <= 0 {
limit = 10 limit = 10
@@ -153,8 +154,9 @@ func (m *Module) RecentSteps(limit int) (lines []StepLine, cut bool, err error)
} }
} }
} }
if len(own) > MaxStateReads { sort.Sort(sort.Reverse(sort.StringSlice(own)))
own, cut = own[:MaxStateReads], true if n := min(max(limit*4, 100), MaxStateReads); len(own) > n {
own, cut = own[:n], true
} }
raws, err := readValues(own, m.ReadSteps.Get) raws, err := readValues(own, m.ReadSteps.Get)
log.Printf("steps read: %d of %d keys listed read in %v (cut %v)", len(own), len(keys), time.Since(start).Round(time.Millisecond), cut) log.Printf("steps read: %d of %d keys listed read in %v (cut %v)", len(own), len(keys), time.Since(start).Round(time.Millisecond), cut)
@@ -197,11 +199,12 @@ func (m *Module) ReadToolchains() (map[string]delivery.Toolchain, error) {
if err != nil { if err != nil {
return nil, err return nil, err
} }
if len(keys) > MaxStateReads { cut := len(keys) > MaxStateReads
if cut {
keys = keys[:MaxStateReads] // a node has a key per language: far fewer in practice keys = keys[:MaxStateReads] // a node has a key per language: far fewer in practice
} }
raws, err := readValues(keys, m.ReadToolchainsState.Get) raws, err := readValues(keys, m.ReadToolchainsState.Get)
log.Printf("toolchains read: %d keys read in %v", len(keys), time.Since(start).Round(time.Millisecond)) log.Printf("toolchains read: %d keys read in %v (cut %v: keys past %d were not read)", len(keys), time.Since(start).Round(time.Millisecond), cut, MaxStateReads)
if err != nil { if err != nil {
return nil, err return nil, err
} }
@@ -242,7 +245,7 @@ func (m *Module) ReadTools() []stdio.Tool {
ans := map[string]any{"node": m.Node, "steps": steps} ans := map[string]any{"node": m.Node, "steps": steps}
if cut { if cut {
ans["cut"] = true ans["cut"] = true
ans["note"] = fmt.Sprintf("this node has more than %d steps stored; %d were read, so the newest may be missing", MaxStateReads, MaxStateReads) ans["note"] = fmt.Sprintf("this node has more steps stored than one call reads; older steps were not read (ask for a larger limit, at most %d reads)", MaxStateReads)
} }
return ans, err return ans, err
}}, }},
@@ -153,14 +153,14 @@ func TestReadToolsAreInTheManifestAndNeedNoToolchain(t *testing.T) {
func TestRecentStepsReadsOnlyTheNodesKeysInParallelAndBounded(t *testing.T) { func TestRecentStepsReadsOnlyTheNodesKeysInParallelAndBounded(t *testing.T) {
t0 := time.Date(2026, 10, 25, 9, 0, 0, 0, time.UTC) t0 := time.Date(2026, 10, 25, 9, 0, 0, 0, time.UTC)
for _, row := range []struct { for _, row := range []struct {
perNode int perNode, limit int
reads int64 reads int64
steps int steps int
cut bool cut bool
}{ }{
{20, 20, 10, false}, {20, 0, 20, 10, false},
{600, 600, 10, false}, {600, 0, 100, 10, true},
{1500, MaxStateReads, 10, true}, {1500, 300, MaxStateReads, 300, true},
} { } {
var reads, inFlight, maxFlight atomic.Int64 var reads, inFlight, maxFlight atomic.Int64
stored := map[string][]byte{} stored := map[string][]byte{}
@@ -193,10 +193,13 @@ func TestRecentStepsReadsOnlyTheNodesKeysInParallelAndBounded(t *testing.T) {
}, },
} }
start := time.Now() start := time.Now()
lines, cut, err := m.RecentSteps(0) lines, cut, err := m.RecentSteps(row.limit)
if err != nil || len(lines) != row.steps || cut != row.cut || reads.Load() != row.reads { if err != nil || len(lines) != row.steps || cut != row.cut || reads.Load() != row.reads {
t.Errorf("%d a node: %d steps, cut %v, %d reads, err %v; want %d, %v, %d", row.perNode, len(lines), cut, reads.Load(), err, row.steps, row.cut, row.reads) t.Errorf("%d a node: %d steps, cut %v, %d reads, err %v; want %d, %v, %d", row.perNode, len(lines), cut, reads.Load(), err, row.steps, row.cut, row.reads)
} }
if row.cut && len(lines) > 0 && lines[0].ID != fmt.Sprintf("s%04d", row.perNode-1) {
t.Errorf("%d a node: newest answered %s", row.perNode, lines[0].ID)
}
if maxFlight.Load() < 2 { if maxFlight.Load() < 2 {
t.Errorf("%d a node: values read one at a time", row.perNode) t.Errorf("%d a node: values read one at a time", row.perNode)
} }
@@ -130,8 +130,9 @@ type StepLine struct {
} }
// RecentSteps are this node's steps, the latest started first, at most limit (ten when limit is not above zero). Only // RecentSteps are this node's steps, the latest started first, at most limit (ten when limit is not above zero). Only
// the node's own keys (<node>.) are read, at most MaxStateReads of them; cut says more were stored than were read, so // the node's own keys (<node>.) are listed and the newest max(limit*4, 100) of them read (at most MaxStateReads): a step's
// the newest may be missing. A step's id says nothing of when it started, so the values are what orders them. // id is base-36 milliseconds and random digits, so its string order is its start order. cut says older steps were not
// read. The values still order the lines, since an ask step's id carries the ask time.
func (m *Module) RecentSteps(limit int) (lines []StepLine, cut bool, err error) { func (m *Module) RecentSteps(limit int) (lines []StepLine, cut bool, err error) {
if limit <= 0 { if limit <= 0 {
limit = 10 limit = 10
@@ -152,8 +153,9 @@ func (m *Module) RecentSteps(limit int) (lines []StepLine, cut bool, err error)
} }
} }
} }
if len(own) > MaxStateReads { sort.Sort(sort.Reverse(sort.StringSlice(own)))
own, cut = own[:MaxStateReads], true if n := min(max(limit*4, 100), MaxStateReads); len(own) > n {
own, cut = own[:n], true
} }
raws, err := readValues(own, m.ReadSteps.Get) raws, err := readValues(own, m.ReadSteps.Get)
log.Printf("steps read: %d of %d keys listed read in %v (cut %v)", len(own), len(keys), time.Since(start).Round(time.Millisecond), cut) log.Printf("steps read: %d of %d keys listed read in %v (cut %v)", len(own), len(keys), time.Since(start).Round(time.Millisecond), cut)
@@ -209,7 +211,7 @@ func (m *Module) ReadTools() []stdio.Tool {
ans := map[string]any{"node": m.Node, "steps": steps} ans := map[string]any{"node": m.Node, "steps": steps}
if cut { if cut {
ans["cut"] = true ans["cut"] = true
ans["note"] = fmt.Sprintf("this node has more than %d steps stored; %d were read, so the newest may be missing", MaxStateReads, MaxStateReads) ans["note"] = fmt.Sprintf("this node has more steps stored than one call reads; older steps were not read (ask for a larger limit, at most %d reads)", MaxStateReads)
} }
return ans, err return ans, err
}}, }},
@@ -54,14 +54,14 @@ func TestTheReadToolsAnswerAStepsResult(t *testing.T) {
func TestRecentStepsReadsOnlyTheNodesKeysInParallelAndBounded(t *testing.T) { func TestRecentStepsReadsOnlyTheNodesKeysInParallelAndBounded(t *testing.T) {
t0 := time.Date(2026, 10, 25, 9, 0, 0, 0, time.UTC) t0 := time.Date(2026, 10, 25, 9, 0, 0, 0, time.UTC)
for _, row := range []struct { for _, row := range []struct {
perNode int perNode, limit int
reads int64 reads int64
steps int steps int
cut bool cut bool
}{ }{
{20, 20, 10, false}, {20, 0, 20, 10, false},
{600, 600, 10, false}, {600, 0, 100, 10, true},
{1500, MaxStateReads, 10, true}, {1500, 300, MaxStateReads, 300, true},
} { } {
var reads, inFlight, maxFlight atomic.Int64 var reads, inFlight, maxFlight atomic.Int64
stored := map[string][]byte{} stored := map[string][]byte{}
@@ -94,10 +94,13 @@ func TestRecentStepsReadsOnlyTheNodesKeysInParallelAndBounded(t *testing.T) {
}, },
} }
start := time.Now() start := time.Now()
lines, cut, err := m.RecentSteps(0) lines, cut, err := m.RecentSteps(row.limit)
if err != nil || len(lines) != row.steps || cut != row.cut || reads.Load() != row.reads { if err != nil || len(lines) != row.steps || cut != row.cut || reads.Load() != row.reads {
t.Errorf("%d a node: %d steps, cut %v, %d reads, err %v; want %d, %v, %d", row.perNode, len(lines), cut, reads.Load(), err, row.steps, row.cut, row.reads) t.Errorf("%d a node: %d steps, cut %v, %d reads, err %v; want %d, %v, %d", row.perNode, len(lines), cut, reads.Load(), err, row.steps, row.cut, row.reads)
} }
if row.cut && len(lines) > 0 && lines[0].ID != fmt.Sprintf("s%04d", row.perNode-1) {
t.Errorf("%d a node: newest answered %s", row.perNode, lines[0].ID)
}
if maxFlight.Load() < 2 { if maxFlight.Load() < 2 {
t.Errorf("%d a node: values read one at a time", row.perNode) t.Errorf("%d a node: values read one at a time", row.perNode)
} }
@@ -131,8 +131,9 @@ type StepLine struct {
} }
// RecentSteps are this node's steps, the latest started first, at most limit (ten when limit is not above zero). Only // RecentSteps are this node's steps, the latest started first, at most limit (ten when limit is not above zero). Only
// the node's own keys (<node>.) are read, at most MaxStateReads of them; cut says more were stored than were read, so // the node's own keys (<node>.) are listed and the newest max(limit*4, 100) of them read (at most MaxStateReads): a step's
// the newest may be missing. A step's id says nothing of when it started, so the values are what orders them. // id is base-36 milliseconds and random digits, so its string order is its start order. cut says older steps were not
// read. The values still order the lines, since an ask step's id carries the ask time.
func (m *Module) RecentSteps(limit int) (lines []StepLine, cut bool, err error) { func (m *Module) RecentSteps(limit int) (lines []StepLine, cut bool, err error) {
if limit <= 0 { if limit <= 0 {
limit = 10 limit = 10
@@ -153,8 +154,9 @@ func (m *Module) RecentSteps(limit int) (lines []StepLine, cut bool, err error)
} }
} }
} }
if len(own) > MaxStateReads { sort.Sort(sort.Reverse(sort.StringSlice(own)))
own, cut = own[:MaxStateReads], true if n := min(max(limit*4, 100), MaxStateReads); len(own) > n {
own, cut = own[:n], true
} }
raws, err := readValues(own, m.ReadSteps.Get) raws, err := readValues(own, m.ReadSteps.Get)
log.Printf("steps read: %d of %d keys listed read in %v (cut %v)", len(own), len(keys), time.Since(start).Round(time.Millisecond), cut) log.Printf("steps read: %d of %d keys listed read in %v (cut %v)", len(own), len(keys), time.Since(start).Round(time.Millisecond), cut)
@@ -197,11 +199,12 @@ func (m *Module) ReadToolchains() (map[string]delivery.Toolchain, error) {
if err != nil { if err != nil {
return nil, err return nil, err
} }
if len(keys) > MaxStateReads { cut := len(keys) > MaxStateReads
if cut {
keys = keys[:MaxStateReads] // a node has a key per language: far fewer in practice keys = keys[:MaxStateReads] // a node has a key per language: far fewer in practice
} }
raws, err := readValues(keys, m.ReadToolchainsState.Get) raws, err := readValues(keys, m.ReadToolchainsState.Get)
log.Printf("toolchains read: %d keys read in %v", len(keys), time.Since(start).Round(time.Millisecond)) log.Printf("toolchains read: %d keys read in %v (cut %v: keys past %d were not read)", len(keys), time.Since(start).Round(time.Millisecond), cut, MaxStateReads)
if err != nil { if err != nil {
return nil, err return nil, err
} }
@@ -242,7 +245,7 @@ func (m *Module) ReadTools() []stdio.Tool {
ans := map[string]any{"node": m.Node, "steps": steps} ans := map[string]any{"node": m.Node, "steps": steps}
if cut { if cut {
ans["cut"] = true ans["cut"] = true
ans["note"] = fmt.Sprintf("this node has more than %d steps stored; %d were read, so the newest may be missing", MaxStateReads, MaxStateReads) ans["note"] = fmt.Sprintf("this node has more steps stored than one call reads; older steps were not read (ask for a larger limit, at most %d reads)", MaxStateReads)
} }
return ans, err return ans, err
}}, }},
@@ -153,14 +153,14 @@ func TestReadToolsAreInTheManifestAndNeedNoToolchain(t *testing.T) {
func TestRecentStepsReadsOnlyTheNodesKeysInParallelAndBounded(t *testing.T) { func TestRecentStepsReadsOnlyTheNodesKeysInParallelAndBounded(t *testing.T) {
t0 := time.Date(2026, 10, 25, 9, 0, 0, 0, time.UTC) t0 := time.Date(2026, 10, 25, 9, 0, 0, 0, time.UTC)
for _, row := range []struct { for _, row := range []struct {
perNode int perNode, limit int
reads int64 reads int64
steps int steps int
cut bool cut bool
}{ }{
{20, 20, 10, false}, {20, 0, 20, 10, false},
{600, 600, 10, false}, {600, 0, 100, 10, true},
{1500, MaxStateReads, 10, true}, {1500, 300, MaxStateReads, 300, true},
} { } {
var reads, inFlight, maxFlight atomic.Int64 var reads, inFlight, maxFlight atomic.Int64
stored := map[string][]byte{} stored := map[string][]byte{}
@@ -193,10 +193,13 @@ func TestRecentStepsReadsOnlyTheNodesKeysInParallelAndBounded(t *testing.T) {
}, },
} }
start := time.Now() start := time.Now()
lines, cut, err := m.RecentSteps(0) lines, cut, err := m.RecentSteps(row.limit)
if err != nil || len(lines) != row.steps || cut != row.cut || reads.Load() != row.reads { if err != nil || len(lines) != row.steps || cut != row.cut || reads.Load() != row.reads {
t.Errorf("%d a node: %d steps, cut %v, %d reads, err %v; want %d, %v, %d", row.perNode, len(lines), cut, reads.Load(), err, row.steps, row.cut, row.reads) t.Errorf("%d a node: %d steps, cut %v, %d reads, err %v; want %d, %v, %d", row.perNode, len(lines), cut, reads.Load(), err, row.steps, row.cut, row.reads)
} }
if row.cut && len(lines) > 0 && lines[0].ID != fmt.Sprintf("s%04d", row.perNode-1) {
t.Errorf("%d a node: newest answered %s", row.perNode, lines[0].ID)
}
if maxFlight.Load() < 2 { if maxFlight.Load() < 2 {
t.Errorf("%d a node: values read one at a time", row.perNode) t.Errorf("%d a node: values read one at a time", row.perNode)
} }