index

package
v1.9.0 Latest Latest
Warning

This package is not in the latest version of its module.

 Go to latest Published: Oct 3, 2024 License: AGPL-3.0 Imports: 27 Imported by: 0

Documentation

Index

Constants

View Source 
const (
	// MagicIndex 4 bytes at the head of an index file.
	MagicIndex = 0xBAAAD700
	// HeaderLen represents number of bytes reserved of index for header.
	HeaderLen = 5

	// FormatV1 represents 1 version of index.
	FormatV1 = 1
	// FormatV2 represents 2 version of index.
	FormatV2 = 2

	IndexFilename = "index"
)
View Source 
const (
	// ShardLabel is a reserved label referencing a cortex shard
	ShardLabel = "__tsdb_shard__"
	// ShardLabelFmt is the fmt of the ShardLabel key.
	ShardLabelFmt = "%d_of_%d"
)

Variables

This section is empty.

Functions

func AllPostingsKey 

func AllPostingsKey() (name, value string)

AllPostingsKey returns the label key that is used to store the postings list of all existing IDs.

func ExpandPostings 

func ExpandPostings(p Postings) (res []storage.SeriesRef, err error)

ExpandPostings returns the postings expanded as a slice.

func ReadOffsetTable 

func ReadOffsetTable(bs ByteSlice, off uint64, f func([]string, uint64, int) error) error

ReadOffsetTable reads an offset table and at the given position calls f for each found entry. If f returns an error it stops decoding and returns the received error.

Types

type ByteSlice 

type ByteSlice interface {
	Len() int
	Range(start, end int) []byte
}

ByteSlice abstracts a byte slice.

type ChunkMeta 

type ChunkMeta struct {
	Checksum uint32

	MinTime, MaxTime int64

	// Bytes stored, rounded to nearest KB
	KB uint32

	SeriesIndex uint32
}

Meta holds information about a chunk of data.

func (ChunkMeta) Bounds 

func (c ChunkMeta) Bounds() (model.Time, model.Time)

func (ChunkMeta) From 

func (c ChunkMeta) From() model.Time

func (ChunkMeta) Through 

func (c ChunkMeta) Through() model.Time

type ChunkMetas 

type ChunkMetas []ChunkMeta

func (ChunkMetas) Bounds 

func (c ChunkMetas) Bounds() (mint, maxt model.Time)

func (ChunkMetas) Finalize 

func (c ChunkMetas) Finalize() ChunkMetas

Finalize sorts and dedupes TODO(owen-d): can we remove the need for this by ensuring we only push in order and without duplicates?

func (ChunkMetas) Len 

func (c ChunkMetas) Len() int

func (ChunkMetas) Less 

func (c ChunkMetas) Less(i, j int) bool

Sort by (MinTime, MaxTime, Checksum)

func (ChunkMetas) Swap 

func (c ChunkMetas) Swap(i, j int)

type Decoder 

type Decoder struct {
	LookupSymbol func(uint32) (string, error)
}

Decoder provides decoding methods for the v1 and v2 index file format.

It currently does not contain decoding methods for all entry types but can be extended by them if there's demand.

func (*Decoder) LabelNamesOffsetsFor 

func (dec *Decoder) LabelNamesOffsetsFor(b []byte) ([]uint32, error)

LabelNamesOffsetsFor decodes the offsets of the name symbols for a given series. They are returned in the same order they're stored, which should be sorted lexicographically.

func (*Decoder) LabelValueFor 

func (dec *Decoder) LabelValueFor(b []byte, label string) (string, error)

LabelValueFor decodes a label for a given series.

func (*Decoder) Postings 

func (dec *Decoder) Postings(b []byte) (int, Postings, error)

Postings returns a postings list for b and its number of elements.

func (*Decoder) Series 

func (dec *Decoder) Series(b []byte, lbls *phlaremodel.Labels, chks *[]ChunkMeta, group bool, by ...string) (uint64, error)

Series decodes a series entry from the given byte slice into lset and chks.

type FileWriter 

type FileWriter struct {
	// contains filtered or unexported fields
}

func NewFileWriter 

func NewFileWriter(name string, bufferSize int) (*FileWriter, error)

func (*FileWriter) AddPadding 

func (fw *FileWriter) AddPadding(size int) error

AddPadding adds zero byte padding until the file size is a multiple size.

func (*FileWriter) Close 

func (fw *FileWriter) Close() error

func (*FileWriter) Flush 

func (fw *FileWriter) Flush() error

func (*FileWriter) Pos 

func (fw *FileWriter) Pos() uint64

func (*FileWriter) Remove 

func (fw *FileWriter) Remove() error

func (*FileWriter) Write 

func (fw *FileWriter) Write(bufs ...[]byte) error

func (*FileWriter) WriteAt 

func (fw *FileWriter) WriteAt(buf []byte, pos uint64) error

type FingerprintOffsets 

type FingerprintOffsets [][2]uint64

(SeriesRef, Fingerprint) tuples

func (FingerprintOffsets) Range 

func (xs FingerprintOffsets) Range(shard ShardAnnotation) (minOffset, maxOffset uint64)

type IndexWriterSeries added in v1.8.0 

type IndexWriterSeries struct {
	Labels phlaremodel.Labels
	Chunks []ChunkMeta // series file offset of chunks
}

type IndexWriterSeriesSlice added in v1.8.0 

type IndexWriterSeriesSlice []*IndexWriterSeries

func (IndexWriterSeriesSlice) Len added in v1.8.0 

func (s IndexWriterSeriesSlice) Len() int

func (IndexWriterSeriesSlice) Less added in v1.8.0 

func (s IndexWriterSeriesSlice) Less(i, j int) bool

func (IndexWriterSeriesSlice) Swap added in v1.8.0 

func (s IndexWriterSeriesSlice) Swap(i, j int)

type MemPostings 

type MemPostings struct {
	// contains filtered or unexported fields
}

MemPostings holds postings list for series ID per label pair. They may be written to out of order. EnsureOrder() must be called once before any reads are done. This allows for quick unordered batch fills on startup.

func NewMemPostings 

func NewMemPostings() *MemPostings

NewMemPostings returns a memPostings that's ready for reads and writes.

func NewUnorderedMemPostings 

func NewUnorderedMemPostings() *MemPostings

NewUnorderedMemPostings returns a memPostings that is not safe to be read from until EnsureOrder() was called once.

func (*MemPostings) Add 

func (p *MemPostings) Add(id storage.SeriesRef, lset phlaremodel.Labels)

Add a label set to the postings index.

func (*MemPostings) All 

func (p *MemPostings) All() Postings

All returns a postings list over all documents ever added.

func (*MemPostings) Delete 

func (p *MemPostings) Delete(deleted map[storage.SeriesRef]struct{})

Delete removes all ids in the given map from the postings lists.

func (*MemPostings) EnsureOrder 

func (p *MemPostings) EnsureOrder()

EnsureOrder ensures that all postings lists are sorted. After it returns all further calls to add and addFor will insert new IDs in a sorted manner.

func (*MemPostings) Get 

func (p *MemPostings) Get(name, value string) Postings

Get returns a postings list for the given label pair.

func (*MemPostings) Iter 

func (p *MemPostings) Iter(f func(labels.Label, Postings) error) error

Iter calls f for each postings list. It aborts if f returns an error and returns it.

func (*MemPostings) LabelNames 

func (p *MemPostings) LabelNames() []string

LabelNames returns all the unique label names.

func (*MemPostings) LabelValues 

func (p *MemPostings) LabelValues(name string) []string

LabelValues returns label values for the given name.

func (*MemPostings) SortedKeys 

func (p *MemPostings) SortedKeys() []labels.Label

SortedKeys returns a list of sorted label keys of the ostings.

func (*MemPostings) Stats 

func (p *MemPostings) Stats(label string) *PostingsStats

Stats calculates the cardinality statistics from postings.

func (*MemPostings) Symbols 

func (p *MemPostings) Symbols() iter.Iterator[string]

Symbols returns an iterator over all unique name and value strings, in order.

type Metadata 

type Metadata struct {
	From, Through int64
	Checksum      uint32
}

Metadata is TSDB-level metadata

func (*Metadata) EnsureBounds 

func (m *Metadata) EnsureBounds(from, through int64)

type PoolChunkMetas 

type PoolChunkMetas struct {
	// contains filtered or unexported fields
}
var ChunkMetasPool PoolChunkMetas

func (*PoolChunkMetas) Get 

func (p *PoolChunkMetas) Get() []ChunkMeta

func (*PoolChunkMetas) Put 

func (p *PoolChunkMetas) Put(xs []ChunkMeta)

type Postings 

type Postings = iter.SeekIterator[storage.SeriesRef, storage.SeriesRef]

Postings provides iterative access over a postings list.

func EmptyPostings 

func EmptyPostings() Postings

EmptyPostings returns a postings list that's always empty. NOTE: Returning EmptyPostings sentinel when Postings struct has no postings is recommended. It triggers optimized flow in other functions like Intersect, Without etc.

func ErrPostings 

func ErrPostings(err error) Postings

ErrPostings returns new postings that immediately error.

func Intersect 

func Intersect(its ...Postings) Postings

Intersect returns a new postings list over the intersection of the input postings.

func Merge 

func Merge(its ...Postings) Postings

Merge returns a new iterator over the union of the input iterators.

func NewBigEndianPostings added in v1.8.0 

func NewBigEndianPostings(list []byte) Postings

func Without 

func Without(full, drop Postings) Postings

Without returns a new postings list that contains all elements from the full list that are not in the drop list.

type PostingsStats 

type PostingsStats struct {
	CardinalityMetricsStats []Stat
	CardinalityLabelStats   []Stat
	LabelValueStats         []Stat
	LabelValuePairsStats    []Stat
	NumLabelPairs           int
}

PostingsStats contains cardinality based statistics for postings.

type Range 

type Range struct {
	Start, End int64
}

Range marks a byte range.

type Reader 

type Reader struct {
	// contains filtered or unexported fields
}

func NewFileReader 

func NewFileReader(path string) (*Reader, error)

NewFileReader returns a new index reader against the given index file.

func NewReader 

func NewReader(b ByteSlice) (*Reader, error)

NewReader returns a new index reader on the given byte slice. It automatically handles different format versions.

func (*Reader) Bounds 

func (r *Reader) Bounds() (int64, int64)

func (*Reader) Checksum 

func (r *Reader) Checksum() uint32

func (*Reader) Close 

func (r *Reader) Close() error

Close the reader and its underlying resources.

func (*Reader) FileInfo 

func (r *Reader) FileInfo() block.File

FileInfo returns some general stats about the underlying file

func (*Reader) LabelNames 

func (r *Reader) LabelNames(matchers ...*labels.Matcher) ([]string, error)

LabelNames returns all the unique label names present in the index. TODO(twilkie) implement support for matchers

func (*Reader) LabelNamesFor 

func (r *Reader) LabelNamesFor(ids ...storage.SeriesRef) ([]string, error)

LabelNamesFor returns all the label names for the series referred to by IDs. The names returned are sorted.

func (*Reader) LabelValueFor 

func (r *Reader) LabelValueFor(id storage.SeriesRef, label string) (string, error)

LabelValueFor returns label value for the given label name in the series referred to by ID.

func (*Reader) LabelValues 

func (r *Reader) LabelValues(name string, matchers ...*labels.Matcher) ([]string, error)

LabelValues returns value tuples that exist for the given label name. It is not safe to use the return value beyond the lifetime of the byte slice passed into the Reader. TODO(replay): Support filtering by matchers

func (*Reader) Postings 

func (r *Reader) Postings(name string, shard *ShardAnnotation, values ...string) (Postings, error)

func (*Reader) PostingsRanges 

func (r *Reader) PostingsRanges() (map[labels.Label]Range, error)

PostingsRanges returns a new map of byte range in the underlying index file for all postings lists.

func (*Reader) Series 

func (r *Reader) Series(id storage.SeriesRef, lbls *phlaremodel.Labels, chks *[]ChunkMeta) (uint64, error)

Series reads the series with the given ID and writes its labels and chunks into lbls and chks.

func (*Reader) SeriesBy added in v1.2.1 

func (r *Reader) SeriesBy(id storage.SeriesRef, lbls *phlaremodel.Labels, chks *[]ChunkMeta, by ...string) (uint64, error)

SeriesBy is like Series but allows to group labels by name. This avoid looking up all label symbols for requested series.

func (*Reader) Size 

func (r *Reader) Size() int64

Size returns the size of an index file.

func (*Reader) SortedLabelValues 

func (r *Reader) SortedLabelValues(name string, matchers ...*labels.Matcher) ([]string, error)

SortedLabelValues returns value tuples that exist for the given label name. It is not safe to use the return value beyond the lifetime of the byte slice passed into the Reader.

func (*Reader) SymbolTableSize 

func (r *Reader) SymbolTableSize() uint64

SymbolTableSize returns the symbol table size in bytes.

func (*Reader) Symbols 

func (r *Reader) Symbols() StringIter

Symbols returns an iterator over the symbols that exist within the index.

func (*Reader) Version 

func (r *Reader) Version() int

Version returns the file format version of the underlying index.

type RealByteSlice 

type RealByteSlice []byte

func (RealByteSlice) Len 

func (b RealByteSlice) Len() int

func (RealByteSlice) Range 

func (b RealByteSlice) Range(start, end int) []byte

func (RealByteSlice) Sub 

func (b RealByteSlice) Sub(start, end int) ByteSlice

type ShardAnnotation 

type ShardAnnotation struct {
	Shard uint32
	Of    uint32
}

ShardAnnotation is a convenience struct which holds data from a parsed shard label Of MUST be a power of 2 to ensure sharding logic works correctly.

func NewShard 

func NewShard(x, of uint32) ShardAnnotation

func (ShardAnnotation) Bounds 

func (shard ShardAnnotation) Bounds() (model.Fingerprint, model.Fingerprint)

Bounds shows the [minimum, maximum) fingerprints. If there is no maximum fingerprint (for example )

func (ShardAnnotation) Match 

func (shard ShardAnnotation) Match(fp model.Fingerprint) bool

Match returns whether a fingerprint belongs to a certain shard. The Shard must be a power of 2. Inclusion in a shard is calculated by determining the arbitrary bit prefix for a shard, then ensuring the fingerprint has the same prefix

func (ShardAnnotation) RequiredBits 

func (shard ShardAnnotation) RequiredBits() uint64

func (ShardAnnotation) String 

func (shard ShardAnnotation) String() string

String encodes a shardAnnotation into a label value

type ShardedPostings 

type ShardedPostings struct {
	// contains filtered or unexported fields
}

func NewShardedPostings 

func NewShardedPostings(p Postings, shard ShardAnnotation, offsets FingerprintOffsets) *ShardedPostings

Note: shardedPostings can technically return more series than just those that correspond to the requested shard. This is because we do fingperint/offset sampling in TSDB so we won't know exactly which offsets to start/end at, but will likely buffer a little on each end, so they still need to be checked for shard inclusiveness. For example (below), given a shard, we'll likely return a slight superset of offsets surrounding the shard. ---[shard0]--- # Shard membership -[--shard0--]- # Series returned by shardedPostings

func (*ShardedPostings) At 

func (sp *ShardedPostings) At() storage.SeriesRef

At returns the value at the current iterator position.

func (*ShardedPostings) Close 

func (sp *ShardedPostings) Close() error

func (*ShardedPostings) Err 

func (sp *ShardedPostings) Err() (err error)

Err returns the last error of the iterator.

func (*ShardedPostings) Next 

func (sp *ShardedPostings) Next() bool

Next advances the iterator and returns true if another value was found.

func (*ShardedPostings) Seek 

func (sp *ShardedPostings) Seek(v storage.SeriesRef) (res bool)

Seek advances the iterator to value v or greater and returns true if a value was found.

type Stat 

type Stat struct {
	Name  string
	Count uint64
}

Stat holds values for a single cardinality statistic.

type StringIter 

type StringIter interface {
	// Next advances the iterator and returns true if another value was found.
	Next() bool

	// At returns the value at the current iterator position.
	At() string

	// Err returns the last error of the iterator.
	Err() error
}

StringIter iterates over a sorted list of strings.

type Symbols 

type Symbols struct {
	// contains filtered or unexported fields
}

func NewSymbols 

func NewSymbols(bs ByteSlice, version, off int) (*Symbols, error)

NewSymbols returns a Symbols object for symbol lookups.

func (Symbols) Iter 

func (s Symbols) Iter() StringIter

func (Symbols) Lookup 

func (s Symbols) Lookup(o uint32) (string, error)

func (Symbols) ReverseLookup 

func (s Symbols) ReverseLookup(sym string) (uint32, error)

func (Symbols) Size 

func (s Symbols) Size() int

type TOC 

type TOC struct {
	Symbols            uint64
	Series             uint64
	LabelIndices       uint64
	LabelIndicesTable  uint64
	Postings           uint64
	PostingsTable      uint64
	FingerprintOffsets uint64
	Metadata           Metadata
}

TOC represents index Table Of Content that states where each section of index starts.

func NewTOCFromByteSlice 

func NewTOCFromByteSlice(bs ByteSlice) (*TOC, error)

NewTOCFromByteSlice return parsed TOC from given index byte slice.

type Writer 

type Writer struct {
	Version int
	// contains filtered or unexported fields
}

Writer implements the IndexWriter interface for the standard serialization format.

func NewWriter 

func NewWriter(ctx context.Context, fn string) (*Writer, error)

NewWriter returns a new Writer to the given filename. It serializes data in format version 2.

func NewWriterSize added in v1.8.0 

func NewWriterSize(ctx context.Context, fn string, bufferSize int) (*Writer, error)

func (*Writer) AddSeries 

func (w *Writer) AddSeries(ref storage.SeriesRef, lset phlaremodel.Labels, fp model.Fingerprint, chunks ...ChunkMeta) error

AddSeries adds the series one at a time along with its chunks. Requires a specific fingerprint to be passed in the case where the "desired" fingerprint differs from what labels.Hash() produces. For example, multitenant TSDBs embed a tenant label, but the actual series has no such label and so the derived fingerprint differs.

func (*Writer) AddSymbol 

func (w *Writer) AddSymbol(sym string) error

func (*Writer) Close 

func (w *Writer) Close() error

Source Files

View all Source files

Jump to

Keyboard shortcuts

? : This menu
/ : Search site
f or F : Jump to
y or Y : Canonical URL
go.dev uses cookies from Google to deliver and enhance the quality of its services and to analyze traffic. Learn more.