Files
KamadoPool/api/internal/logmon/tailer.go
T
satoshi df0dbf89e5 Harden block-recording pipeline: P0 reliability fixes
Closes the silent-failure modes between "ckpool logs a solve" and
"block correctly displayed":

* Difficulty estimate matched mempool.space — the projection now uses
  (inEpoch + 1) intervals so it converges on Bitcoin Core's eventual
  retarget formula at end-of-epoch instead of undershooting by ~0.05–
  0.10 % throughout.

* Tailer resumes mid-log on restart — persists (inode, offset) to kv
  every EOF + on shutdown, and replays the unread tail next time. Any
  solve line written while kamado-api was down would previously be
  invisible forever.

* Background reconcile loop (60 s) retries hash/reward enrichment for
  blocks the original RPC missed, so a transient bitcoind-index race no
  longer permanently leaves a block hashless.

* Reorg detection: same loop compares each recent stored hash against
  getblockhash(height); a mismatch stamps orphaned_at. UI renders these
  strikethrough with a red "orphaned" tag instead of showing illusory
  rewards forever.

* InsertBlock now reports whether a row was actually inserted; the
  caller WARN-logs duplicate-height ignores so a re-mined orphaned
  height can't disappear silently.

* Submit-attempt vs confirmed counters surface failed submissions:
  every "Possible/Submitting block solve" log line increments
  block_submit_attempts; "Solved and confirmed" increments
  block_submits_confirmed. A growing gap means bitcoind is rejecting
  our submissions — previously invisible.

* share_err patch refreshed against pinned ckpool source: added
  SE_NO_JOBID -> 21 and SE_WORKER_MISMATCH -> 24 mappings, kept
  SE_INVALID_NONCE2 in 20 (it's a malformed-input error, not low-diff).
  AxeOS users now see actionable Stratum codes instead of
  "unknown error".

UI gets new orphaned_at + block_submit_attempts/confirmed fields on
the snapshot type and a strikethrough-with-tag rendering for orphaned
blocks in BlocksTable.
2026-04-27 16:30:15 +03:00

289 lines
7.7 KiB
Go

// Package logmon tails the ckpool log file looking for notable events —
// primarily block-solve lines, which are the most reliable signal we have
// that a block was found, short of a ZMQ hashblock subscription.
//
// ckpool-solo logs a line like:
//
// Solved and confirmed block 840123
//
// from stratifier.c via LOGWARNING when a submitted share passes network
// difficulty and bitcoind confirms acceptance. We parse these lines,
// emit BlockEvent values on Events, and let the aggregator enrich them
// with hash/reward via bitcoind RPC.
package logmon
import (
"bufio"
"context"
"errors"
"io"
"log/slog"
"os"
"regexp"
"strconv"
"time"
)
// BlockEvent is emitted when the tailer sees a "Solved and confirmed block"
// line in the ckpool log. Hash and worker are populated later by the
// aggregator once it cross-references bitcoind. ShareDiff is the
// difficulty of the winning share, captured from the "Possible block
// solve" line that precedes the confirmation line.
type BlockEvent struct {
Height int64 `json:"height"`
SeenAt time.Time `json:"seen_at"`
RawLine string `json:"raw_line"`
ShareDiff float64 `json:"share_diff,omitempty"`
}
// AttemptEvent is emitted when ckpool logs a "Possible block solve"
// or "Submitting possible block solve" line — i.e. ckpool decided a
// share met network difficulty and is attempting to submit it to
// bitcoind. This precedes the "Solved and confirmed" line that drives
// BlockEvent. Counting attempts vs confirmations surfaces submission
// failures (bitcoind rejected, RPC timeout, etc.) that would otherwise
// be invisible.
type AttemptEvent struct {
SeenAt time.Time
ShareDiff float64
RawLine string
}
// Tailer follows a log file, surviving rotation/truncation, and emits
// parsed events. Create with New, then Run in a goroutine.
type Tailer struct {
Path string
Events chan BlockEvent
Attempts chan AttemptEvent
Log *slog.Logger
PollWait time.Duration // how long to sleep between EOF polls
// LoadCursor / SaveCursor, if both non-nil, persist the read
// position across process restarts. LoadCursor returns (inode,
// offset, true) when a saved cursor exists for this path; the
// tailer will resume from offset only if the inode still matches.
// SaveCursor is invoked on a throttled cadence as we read, so a
// crash loses at most ~1s of unread bytes.
LoadCursor func() (inode uint64, offset int64, ok bool)
SaveCursor func(inode uint64, offset int64)
// lastSolveDiff remembers the share diff from the most recent
// "Possible block solve" line so handleLine can attach it to the
// subsequent "Solved and confirmed block" event. Reset after use.
lastSolveDiff float64
lastCursorSave time.Time
}
func New(path string, log *slog.Logger) *Tailer {
return &Tailer{
Path: path,
Events: make(chan BlockEvent, 16),
Attempts: make(chan AttemptEvent, 16),
Log: log,
PollWait: 500 * time.Millisecond,
}
}
var (
solvedRE = regexp.MustCompile(`Solved and confirmed block\s+(\d+)`)
// Matches the three "Possible ... block solve ... diff <float>" lines
// ckpool emits from stratifier.c right before a block is submitted:
// "Possible block solve diff N !"
// "Possible stale share block solve diff N !"
// "Submitting possible block solve share diff N !"
// "Possible remote block solve diff N !"
solveDiffRE = regexp.MustCompile(`(?:Possible|Submitting[^"]*possible).*block solve.*diff\s+([0-9eE.+-]+)`)
)
// Run blocks until ctx is cancelled. It opens the file, seeks to either
// the saved cursor (if LoadCursor returns one and the inode still
// matches) or the end on first-ever run, and reads new lines as they
// are appended. If the file is rotated (shrinks, or inode changes), it
// re-opens.
func (t *Tailer) Run(ctx context.Context) {
defer close(t.Events)
defer close(t.Attempts)
var (
f *os.File
reader *bufio.Reader
lastIno uint64
lastPos int64
// firstOpen distinguishes the very first Open of this tailer
// (where we honor LoadCursor and replay the backlog) from
// rotation re-opens (where we always start at 0).
firstOpen = true
)
open := func() error {
if f != nil {
_ = f.Close()
}
nf, err := os.Open(t.Path)
if err != nil {
return err
}
st, err := nf.Stat()
if err != nil {
_ = nf.Close()
return err
}
ino := inodeOf(st)
size := st.Size()
// Decide where to start reading.
var startAt int64
if firstOpen {
if t.LoadCursor != nil {
if savedIno, savedOff, ok := t.LoadCursor(); ok &&
savedIno == ino && savedOff <= size {
startAt = savedOff
if savedOff < size {
t.Log.Info("logmon: resuming from saved cursor",
"path", t.Path, "offset", savedOff, "size", size)
}
} else {
// No cursor, or stale (file was rotated since we
// last saw it, so we have no way to know what we
// already read). Start at end to avoid replaying
// the entire log.
startAt = size
}
} else {
startAt = size
}
firstOpen = false
} else {
// Rotation: read the whole replacement file from the start.
startAt = 0
}
if _, err := nf.Seek(startAt, io.SeekStart); err != nil {
_ = nf.Close()
return err
}
f = nf
reader = bufio.NewReader(f)
lastIno = ino
lastPos = startAt
return nil
}
saveCursor := func(force bool) {
if t.SaveCursor == nil || f == nil {
return
}
now := time.Now()
if !force && now.Sub(t.lastCursorSave) < time.Second {
return
}
t.SaveCursor(lastIno, lastPos)
t.lastCursorSave = now
}
// Initial open; retry on failure until the file exists.
for {
if err := open(); err != nil {
t.Log.Warn("logmon: waiting for log file", "path", t.Path, "err", err)
if !sleep(ctx, 2*time.Second) {
return
}
continue
}
break
}
defer func() {
if f != nil {
_ = f.Close()
}
}()
for {
if ctx.Err() != nil {
saveCursor(true)
return
}
line, err := reader.ReadString('\n')
if len(line) > 0 {
t.handleLine(line)
lastPos, _ = f.Seek(0, io.SeekCurrent)
}
if err == nil {
continue
}
if !errors.Is(err, io.EOF) {
t.Log.Warn("logmon: read error, reopening", "err", err)
saveCursor(true)
if !sleep(ctx, t.PollWait) {
return
}
_ = open()
continue
}
// EOF: persist where we are (caught up to current end), then
// check for rotation (inode changed) or truncation (size < pos).
saveCursor(false)
if st, statErr := os.Stat(t.Path); statErr == nil {
if inodeOf(st) != lastIno || st.Size() < lastPos {
t.Log.Info("logmon: log rotated, reopening", "path", t.Path)
if err := open(); err != nil {
t.Log.Warn("logmon: reopen failed", "err", err)
}
continue
}
}
if !sleep(ctx, t.PollWait) {
return
}
}
}
func (t *Tailer) handleLine(line string) {
if m := solveDiffRE.FindStringSubmatch(line); m != nil {
var d float64
if v, err := strconv.ParseFloat(m[1], 64); err == nil {
d = v
t.lastSolveDiff = v
}
select {
case t.Attempts <- AttemptEvent{SeenAt: time.Now(), ShareDiff: d, RawLine: line}:
default:
// Best-effort metric — dropping is fine.
}
return
}
m := solvedRE.FindStringSubmatch(line)
if m == nil {
return
}
height, err := strconv.ParseInt(m[1], 10, 64)
if err != nil {
return
}
ev := BlockEvent{
Height: height,
SeenAt: time.Now(),
RawLine: line,
ShareDiff: t.lastSolveDiff,
}
t.lastSolveDiff = 0
select {
case t.Events <- ev:
t.Log.Info("logmon: block solved", "height", height, "share_diff", ev.ShareDiff)
default:
t.Log.Warn("logmon: events channel full, dropping", "height", height)
}
}
// sleep returns false if ctx was cancelled during the wait.
func sleep(ctx context.Context, d time.Duration) bool {
select {
case <-ctx.Done():
return false
case <-time.After(d):
return true
}
}