VictoriaMetrics/lib/storage/part.go
Aliaksandr Valialkin 43b24164ef
all: add Windows build for VictoriaMetrics
This commit changes background merge algorithm, so it becomes compatible with Windows file semantics.

The previous algorithm for background merge:

1. Merge source parts into a destination part inside tmp directory.
2. Create a file in txn directory with instructions on how to atomically
   swap source parts with the destination part.
3. Perform instructions from the file.
4. Delete the file with instructions.

This algorithm guarantees that either source parts or destination part
is visible in the partition after unclean shutdown at any step above,
since the remaining files with instructions is replayed on the next restart,
after that the remaining contents of the tmp directory is deleted.

Unfortunately this algorithm doesn't work under Windows because
it disallows removing and moving files, which are in use.

So the new algorithm for background merge has been implemented:

1. Merge source parts into a destination part inside the partition directory itself.
   E.g. now the partition directory may contain both complete and incomplete parts.
2. Atomically update the parts.json file with the new list of parts after the merge,
   e.g. remove the source parts from the list and add the destination part to the list
   before storing it to parts.json file.
3. Remove the source parts from disk when they are no longer used.

This algorithm guarantees that either source parts or destination part
is visible in the partition after unclean shutdown at any step above,
since incomplete partitions from step 1 or old source parts from step 3 are removed
on the next startup by inspecting parts.json file.

This algorithm should work under Windows, since it doesn't remove or move files in use.
This algorithm has also the following benefits:

- It should work better for NFS.
- It fits object storage semantics.

The new algorithm changes data storage format, so it is impossible to downgrade
to the previous versions of VictoriaMetrics after upgrading to this algorithm.

Updates https://github.com/VictoriaMetrics/VictoriaMetrics/issues/3236
Updates https://github.com/VictoriaMetrics/VictoriaMetrics/issues/3821
Updates https://github.com/VictoriaMetrics/VictoriaMetrics/issues/70
2023-03-19 01:36:51 -07:00

138 lines
3.5 KiB
Go

package storage
import (
"fmt"
"path/filepath"
"sync"
"unsafe"
"github.com/VictoriaMetrics/VictoriaMetrics/lib/blockcache"
"github.com/VictoriaMetrics/VictoriaMetrics/lib/filestream"
"github.com/VictoriaMetrics/VictoriaMetrics/lib/fs"
"github.com/VictoriaMetrics/VictoriaMetrics/lib/memory"
)
var ibCache = blockcache.NewCache(getMaxIndexBlocksCacheSize)
func getMaxIndexBlocksCacheSize() int {
maxIndexBlockCacheSizeOnce.Do(func() {
maxIndexBlockCacheSize = int(0.1 * float64(memory.Allowed()))
})
return maxIndexBlockCacheSize
}
var (
maxIndexBlockCacheSize int
maxIndexBlockCacheSizeOnce sync.Once
)
// part represents a searchable part containing time series data.
type part struct {
ph partHeader
// Filesystem path to the part.
//
// Empty for in-memory part.
path string
// Total size in bytes of part data.
size uint64
timestampsFile fs.MustReadAtCloser
valuesFile fs.MustReadAtCloser
indexFile fs.MustReadAtCloser
metaindex []metaindexRow
}
// openFilePart opens file-based part from the given path.
func openFilePart(path string) (*part, error) {
path = filepath.Clean(path)
var ph partHeader
if err := ph.ReadMetadata(path); err != nil {
return nil, fmt.Errorf("cannot parse path to part: %w", err)
}
timestampsPath := path + "/timestamps.bin"
timestampsFile := fs.MustOpenReaderAt(timestampsPath)
timestampsSize := fs.MustFileSize(timestampsPath)
valuesPath := path + "/values.bin"
valuesFile := fs.MustOpenReaderAt(valuesPath)
valuesSize := fs.MustFileSize(valuesPath)
indexPath := path + "/index.bin"
indexFile := fs.MustOpenReaderAt(indexPath)
indexSize := fs.MustFileSize(indexPath)
metaindexPath := path + "/metaindex.bin"
metaindexFile, err := filestream.Open(metaindexPath, true)
if err != nil {
timestampsFile.MustClose()
valuesFile.MustClose()
indexFile.MustClose()
return nil, fmt.Errorf("cannot open metaindex file: %w", err)
}
metaindexSize := fs.MustFileSize(metaindexPath)
size := timestampsSize + valuesSize + indexSize + metaindexSize
return newPart(&ph, path, size, metaindexFile, timestampsFile, valuesFile, indexFile)
}
// newPart returns new part initialized with the given arguments.
//
// The returned part calls MustClose on all the files passed to newPart
// when calling part.MustClose.
func newPart(ph *partHeader, path string, size uint64, metaindexReader filestream.ReadCloser, timestampsFile, valuesFile, indexFile fs.MustReadAtCloser) (*part, error) {
var errors []error
metaindex, err := unmarshalMetaindexRows(nil, metaindexReader)
if err != nil {
errors = append(errors, fmt.Errorf("cannot unmarshal metaindex data: %w", err))
}
metaindexReader.MustClose()
var p part
p.ph = *ph
p.path = path
p.size = size
p.timestampsFile = timestampsFile
p.valuesFile = valuesFile
p.indexFile = indexFile
p.metaindex = metaindex
if len(errors) > 0 {
// Return only the first error, since it has no sense in returning all errors.
err = fmt.Errorf("cannot initialize part %q: %w", &p, errors[0])
p.MustClose()
return nil, err
}
return &p, nil
}
// String returns human-readable representation of p.
func (p *part) String() string {
if len(p.path) > 0 {
return p.path
}
return p.ph.String()
}
// MustClose closes all the part files.
func (p *part) MustClose() {
p.timestampsFile.MustClose()
p.valuesFile.MustClose()
p.indexFile.MustClose()
ibCache.RemoveBlocksForPart(p)
}
type indexBlock struct {
bhs []blockHeader
}
func (idxb *indexBlock) SizeBytes() int {
return cap(idxb.bhs) * int(unsafe.Sizeof(blockHeader{}))
}