wal: initial interfaces for wal package

Informs cockroachdb#3230

wal/wal.go

@@ -0,0 +1,185 @@
+// Copyright 2024 The LevelDB-Go and Pebble Authors. All rights reserved. Use
+// of this source code is governed by a BSD-style license that can be found in
+// the LICENSE file.
+
+package wal
+
+import (
+	"fmt"
+	"io"
+	"sync"
+	"time"
+
+	"github.com/cockroachdb/pebble/internal/base"
+	"github.com/cockroachdb/pebble/record"
+	"github.com/cockroachdb/pebble/vfs"
+	"github.com/prometheus/client_golang/prometheus"
+)
+
+// TODO(sumeer): write a high-level comment describing the approach.
+
+// Dir is used for storing log files.
+type Dir struct {
+	FS      vfs.FS
+	Dirname string
+}
+
+// NumWAL is the number of the virtual WAL. It can map to one or more physical
+// log files. In standalone mode, it will map to exactly one log file. In
+// failover mode, it can map to many log files, which are totally ordered
+// (using a dense logIndex).
+//
+// In general, WAL refers to the virtual WAL, and file refers to a log file.
+// The Pebble MANIFEST only knows about virtual WALs and assigns numbers to
+// them. Additional mapping to one or more files happens in this package. If a
+// WAL maps to multiple files, the source of truth regarding that mapping is
+// the contents of the directories.
+type NumWAL base.DiskFileNum
+
+// logIndex numbers log files within a WAL.
+type logIndex uint32
+
+// TODO(sumeer): parsing func. And remove attempts to parse log files outside
+// the wal package (including tools).
+
+// makeLogFilename makes a log filename.
+func makeLogFilename(wn NumWAL, index logIndex) string {
+	if index == 0 {
+		// Use a backward compatible name, for simplicity.
+		return base.MakeFilename(base.FileTypeLog, base.DiskFileNum(wn))
+	}
+	return fmt.Sprintf("%s-%03d.log", base.DiskFileNum(wn).String(), index)
+}
+
+// Options provides configuration for the Manager.
+type Options struct {
+	// Primary dir for storing WAL files.
+	Primary Dir
+	// Secondary is used for failover. Optional.
+	Secondary Dir
+
+	// Recyling configuration. Only files in the primary dir are recycled.
+
+	// MinRecycleLogNum is the minimum log file number that is allowed to be
+	// recycled. Log file numbers smaller than this will be deleted. This is
+	// used to prevent recycling a log written by a previous instance of the DB
+	// which may not have had log recycling enabled.
+	MinRecycleLogNum base.DiskFileNum
+	// MaxNumRecyclableLogs is the maximum number of log files to maintain for
+	// recycling.
+	MaxNumRecyclableLogs int
+
+	// SyncingFileOptions is the configuration when calling vfs.NewSyncingFile.
+	SyncingFileOpts vfs.SyncingFileOptions
+
+	// MinSyncInterval is documented in Options.WALMinSyncInterval.
+	MinSyncInterval func() time.Duration
+	// FsyncLatency records fsync latency. This doesn't differentiate between
+	// fsyncs on the primary and secondary dir.
+	//
+	// TODO(sumeer): consider separating out into two histograms.
+	FsyncLatency prometheus.Histogram
+	// QueueSemChan is the channel to pop from when popping from queued records
+	// that have requested a sync. It's original purpose was to function as a
+	// semaphore that prevents the record.LogWriter.flusher.syncQueue from
+	// overflowing (which will cause a panic). It is still useful in that role
+	// when the WALManager is configured in standalone mode. In failover mode
+	// there is no syncQueue, so the pushback into the commit pipeline is
+	// unnecessary, but possibly harmless.
+	QueueSemChan chan struct{}
+}
+
+// Stats exposes stats used in Pebble metrics.
+type Stats struct {
+	// ObsoleteFileCount is the number of obsolete log files.
+	ObsoleteFileCount int
+	// ObsoleteFileSize is the total size of obsolete log files.
+	ObsoleteFileSize uint64
+	// LiveFileCount is the number of live log files.
+	LiveFileCount int
+	// LiveFileSize is the total size of live log files. This can be higher than
+	// LiveSize due to log recycling (a live log file may be larger than the
+	// size used in its latest incarnation), or failover (resulting in multiple
+	// log files containing the same records).
+	LiveFileSize uint64
+	// LiveSize is the total size of the live data in log files.
+	LiveSize uint64
+}
+
+// Manager handles all WAL work.
+//
+// It is an interface for now, but if we end up with a single implementation
+// for both standalone mode and failover mode, we will get rid of the
+// interface.
+type Manager interface {
+	// Init initializes the Manager.
+	//
+	// Implementation notes:
+	// - lists and stats the directories, so that Stats are up to date (assuming
+	//   no obsolete files yet), and the list of WALs and their constituent log
+	//   files is initialized.
+	// - ensures dirs are created and synced.
+	Init(o Options) error
+	// List returns the virtual WALs in ascending order.
+	List() ([]NumWAL, error)
+	// Delete deletes all virtual WALs up to highestObsoleteNum. The
+	// underlying physical WAL files may be recycled.
+	Delete(highestObsoleteNum NumWAL) error
+	// OpenForRead opens a virtual WAL for read.
+	OpenForRead(wn NumWAL, strictWALTail bool) (Reader, error)
+	// Create creates a new virtual WAL.
+	//
+	// NumWALs passed to successive Create calls must be monotonically
+	// increasing, and be greater than any NumWAL seen earlier. The caller must
+	// close the previous Writer before calling Create.
+	Create(wn NumWAL) (Writer, error)
+	// Stats returns the latest Stats.
+	Stats() Stats
+	// Close the manager.
+	Close() error
+}
+
+// SyncOptions has non-nil Done and Err when fsync is requested, else both are
+// nil.
+type SyncOptions struct {
+	Done *sync.WaitGroup
+	Err  *error
+}
+
+// Writer writes to a virtual WAL. A Writer in standalone mode maps to a
+// single record.LogWriter. In failover mode, it can failover across multiple
+// physical log files.
+type Writer interface {
+	// Size based on writes.
+	Size() uint64
+	// FileSize is the size of the file(s) underlying this WAL. FileSize
+	// >= Size because of recycling and failover. This is an estimate.
+	FileSize() uint64
+	// WriteRecord writes a complete record. The record is asynchronously
+	// persisted to the underlying writer. If SyncOptions.Done != nil, the wait
+	// group will be notified when durability is guaranteed or an error has
+	// occurred (set in SyncOptions.Err). External synchronisation provided by
+	// commitPipeline.mu guarantees that WriteRecord calls are serialized.
+	WriteRecord(p []byte, opts SyncOptions) error
+	// Close the writer.
+	Close() error
+	// Metrics must be called after Close. The callee will no longer modify the
+	// returned LogWriterMetrics.
+	Metrics() *record.LogWriterMetrics
+}
+
+// Reader reads a virtual WAL.
+type Reader interface {
+	// NextRecord returns the next record, or error.
+	NextRecord() (io.Reader, error)
+	// LogicalOffset is the monotonically increasing offset in the WAL. When the
+	// WAL corresponds to a single log file, this is the offset in that log
+	// file.
+	LogicalOffset() int64
+	// Close the reader.
+	Close() error
+}
+
+// Make lint happy.
+var _ logIndex = 0
+var _ = makeLogFilename