InfluxDB -- TSM存储引擎的读写操作

a朋

数据写入

数据写入时,首先points按shard划分,归属于一个shard的points一起写入:

//tsdb/store.go
// WriteToShard writes a list of points to a shard identified by its ID.
func (s *Store) WriteToShard(shardID uint64, points []models.Point) error {
    sh := s.shards[shardID]
    return sh.WritePoints(points)
}
//tsdb/shard.go
// WritePoints will write the raw data points and any new metadata to the index in the shard.
func (s *Shard) WritePoints(points []models.Point) error {
    .....
    // Write to the engine.
    err := engine.WritePoints(points);
    .....
}

由tsm1.Engine负责写入points:

  • 首先,构造数据,由points构造values=map[string][]Values,key=seriesKey+分隔符+fieldName, value=[]Value={timestamp,fieldValue}集合;
  • 然后,将values写入cache;
  • 最后,将values写入WAL;
//tsdb/engine/tsm1/engine.go
// WritePoints writes metadata and point data into the engine.
// It returns an error if new points are added to an existing key.
func (e *Engine) WritePoints(points []models.Point) error {
    values := make(map[string][]Value, len(points))
    for _, p := range points {
        keyBuf = append(keyBuf[:0], p.Key()...)
        keyBuf = append(keyBuf, keyFieldSeparator...)
        //一个Point中可能含多个field
        iter := p.FieldIterator()
        t := p.Time().UnixNano()
        for iter.Next() {
            keyBuf = append(keyBuf[:baseLen], iter.FieldKey()...)
            var v Value
            switch iter.Type() {
            case models.Float:
                fv, err := iter.FloatValue()
                if err != nil {
                    return err
                }
                v = NewFloatValue(t, fv)
                ......
            }
            values[string(keyBuf)] = append(values[string(keyBuf)], v)
        }
    }
    //先写到cache
    // first try to write to the cache
    if err := e.Cache.WriteMulti(values); err != nil {
        return err
    }
    //再写到WAL
    if e.WALEnabled {
        if _, err := e.WAL.WriteMulti(values); err != nil {
            return err
        }
    }
    return seriesErr
}

数据删除

与LSM-Tree类似,influxdb使用标记删除的方法,待执行compactor的时候,再真正的将其删除。
在data目录,有.tombstone文件,记录了哪个时间段的数据需要删除:

  • 查询时,将查询结果和.tombstone内容比对,将要删除的记录去掉;
  • compactor时,查询.tombstone内容,将数据删除;

数据查询与索引结构

LSM-Tree有良好的写入性能,但是查询性能不足;TSM-Tree基于LSM-Tree,通过采用索引、布隆过滤器的方法进行查询优化,这里重点关注索引。
influxdb中有两种类型的索引:元数据索引和TSM File索引

元数据索引

元数据指measurement和series信息,每个database都有一个Index结构,存储该database中的元数据索引信息:

//tsdb/store.go
type Store struct {
    path              string
    // shared per-database indexes, only if using "inmem".
    indexes map[string]interface{}    //key=databaseName, value实际是*Index
    ....
}

元数据索引的内部结构:

type Index struct {
    //数据库下name-->*measurement
    measurements map[string]*measurement // measurement name to object and index
    //数据库下seriesKey-->*series
    series       map[string]*series      // map series key to the Series object
    //数据库名称
    database string
}
type measurement struct {
    Database  string
    Name      string `json:"name,omitempty"`
    fieldNames map[string]struct{}
    // in-memory index fields
    //seriesId-->*series
    seriesByID          map[uint64]*series      // lookup table for series by their id
    //tagKey-->tagValue-->[]seriesId
    //查询时,可根据tagKey找到seriesId,然后再找到相关的series
    seriesByTagKeyValue map[string]*tagKeyValue // map from tag key to value to sorted set of series ids
    sortedSeriesIDs seriesIDs // sorted list of series IDs in this measurement
}
type tagKeyValue struct {
    mu      sync.RWMutex
    entries map[string]*tagKeyValueEntry
}
type tagKeyValueEntry struct {
    m map[uint64]struct{} // series id set
}

Image.png

对于元数据查询语句:

show tag values from "cpu_usage" with key="host"

该语句的查询过程:

  • 根据"cpu_usage"找到measurement对象;
  • 在measurement对象内,根据tagKey="host",找到其对应的tagValue+[]seriesId;

对于普通查询语句:

select value from "cpu_usage" where host='server01' and time > now() - 1h

该语句的查询过程:

  • 根据时间:time > now() - 1h,得到数据shard;
  • 在shard内,根据"cpu_usage"找到measurement对象;
  • 在measurement对象内,根据tagKey="server01",找到其对应的tagValue+[]seriesId;
  • 遍历[]seriesId,获得[]series对象,再使用TSM File索引查找TSM File,读取TSM File block得到结果;

TSM File索引

单个TSM File中包含block数据和index数据:

image.png

Blocks中存放压缩后的timestamp/value。
Index中存放Block中的索引,Index会存储到内存做间接索引,以便实现快速检索。

间接索引的数据结构:

//tsdb/engine/tsm1/reader.go
type indirectIndex struct {
    b []byte    //Index的内容
    offsets []byte
    minKey, maxKey []byte    //最小/最大key
    minTime, maxTime int64    //最小/最大时间
}

image.png

TSM File的查找过程:

  • 根据seriesKey,在[]offsetIndex中各offset的key进行二分查找,得到offset;
  • 根据offset读取[]byte内容,得到indexEntries;
  • 在indexEntries中,得到TSM File的偏移量,然后读取文件内容得到结果;
//tsdb/engine/tsm1/reader.go
type indexEntries struct {
    Type    byte
    entries []IndexEntry
}
// IndexEntry is the index information for a given block in a TSM file.
type IndexEntry struct {
    // The min and max time of all points stored in the block.
    MinTime, MaxTime int64
    // The absolute position in the file where this block is located.
    Offset int64    //TSM文件的偏移量
    // The size in bytes of the block in the file.
    Size uint32
}

参考

1.http://blog.fatedier.com/2016...

阅读 190
10 声望
5 粉丝
0 条评论
10 声望
5 粉丝
文章目录
宣传栏