Documentation ¶
Index ¶
- Constants
- Variables
- func GetMaxTidRecordForTest() int64
- type Handle
- func (h *Handle) Clear()
- func (h *Handle) Close()
- func (h *Handle) FlushStats()
- func (h *Handle) GetPartitionStats(tblInfo *model.TableInfo, pid int64) *statistics.Table
- func (h *Handle) GetPartitionStatsForAutoAnalyze(tblInfo *model.TableInfo, pid int64) *statistics.Table
- func (h *Handle) GetTableStats(tblInfo *model.TableInfo) *statistics.Table
- func (h *Handle) GetTableStatsForAutoAnalyze(tblInfo *model.TableInfo) *statistics.Table
- func (h *Handle) InitStats(ctx context.Context, is infoschema.InfoSchema) (err error)
- func (h *Handle) InitStatsLite(ctx context.Context) (err error)
- func (h *Handle) StartWorker()
- type MaxTidRecord
Constants ¶
const ( // StatsOwnerKey is the stats owner path that is saved to etcd. StatsOwnerKey = "/tidb/stats/owner" // StatsPrompt is the prompt for stats owner manager. StatsPrompt = "stats" )
Variables ¶
var IsFullCacheFunc func(cache statstypes.StatsCache, total uint64) bool = isFullCache
IsFullCacheFunc is whether the cache is full or not. but we can only change it when to test
Functions ¶
func GetMaxTidRecordForTest ¶
func GetMaxTidRecordForTest() int64
GetMaxTidRecordForTest gets the max tid record for test.
Types ¶
type Handle ¶
type Handle struct { // Pool is used to get a session or a goroutine to execute stats updating. util.Pool // AutoAnalyzeProcIDGenerator is used to generate auto analyze proc ID. util.AutoAnalyzeProcIDGenerator // LeaseGetter is used to get stats lease. util.LeaseGetter // TableInfoGetter is used to fetch table meta info. util.TableInfoGetter // StatsGC is used to GC stats. types.StatsGC // StatsUsage is used to track the usage of column / index statistics. types.StatsUsage // StatsHistory is used to manage historical stats. types.StatsHistory // StatsAnalyze is used to handle auto-analyze and manage analyze jobs. types.StatsAnalyze // StatsSyncLoad is used to load stats syncly. types.StatsSyncLoad // StatsReadWriter is used to read/write stats from/to storage. types.StatsReadWriter // StatsLock is used to manage locked stats. types.StatsLock // StatsGlobal is used to manage global stats. types.StatsGlobal // DDL is used to handle ddl events. types.DDL InitStatsDone chan struct{} // StatsCache ... types.StatsCache // contains filtered or unexported fields }
Handle can update stats info periodically.
func NewHandle ¶
func NewHandle( _, initStatsCtx sessionctx.Context, lease time.Duration, is infoschema.InfoSchema, pool pkgutil.SessionPool, tracker sysproctrack.Tracker, ddlNotifier *notifier.DDLNotifier, autoAnalyzeProcIDGetter func() uint64, releaseAutoAnalyzeProcID func(uint64), ) (*Handle, error)
NewHandle creates a Handle for update stats.
func (*Handle) FlushStats ¶
func (h *Handle) FlushStats()
FlushStats flushes the cached stats update into store.
func (*Handle) GetPartitionStats ¶
GetPartitionStats retrieves the partition stats from cache. TODO: remove GetTableStats later on.
func (*Handle) GetPartitionStatsForAutoAnalyze ¶
func (h *Handle) GetPartitionStatsForAutoAnalyze(tblInfo *model.TableInfo, pid int64) *statistics.Table
GetPartitionStatsForAutoAnalyze is to get partition stats but it will not return pseudo stats.
func (*Handle) GetTableStats ¶
func (h *Handle) GetTableStats(tblInfo *model.TableInfo) *statistics.Table
GetTableStats retrieves the statistics table from cache, and the cache will be updated by a goroutine. TODO: remove GetTableStats later on.
func (*Handle) GetTableStatsForAutoAnalyze ¶
func (h *Handle) GetTableStatsForAutoAnalyze(tblInfo *model.TableInfo) *statistics.Table
GetTableStatsForAutoAnalyze is to get table stats but it will not return pseudo stats.
func (*Handle) InitStats ¶
func (h *Handle) InitStats(ctx context.Context, is infoschema.InfoSchema) (err error)
InitStats initiates the stats cache. 1. Basic stats meta data is loaded.(count, modify count, etc.) 2. Index stats are fully loaded. (histogram, topn, buckets) 2. Column stats are marked as existing or not by initializing the table.ColAndIdxExistenceMap, based on data from mysql.stats_histograms) To work with auto-analyze's needs, we need to read all stats meta info into memory. The sync/async load of the stats or other process haven't done a full initialization of the table.ColAndIdxExistenceMap. So we need to it here.
func (*Handle) InitStatsLite ¶
InitStatsLite initiates the stats cache. The function is liter and faster than InitStats. 1. Basic stats meta data is loaded.(count, modify count, etc.) 2. Column/index stats are marked as existing or not by initializing the table.ColAndIdxExistenceMap, based on data from mysql.stats_histograms) 3. TopN, Bucket, FMSketch are not loaded. And to work with auto analyze's needs, we need to read all the tables' stats meta into memory. The sync/async load of the stats or other process haven't done a full initialization of the table.ColAndIdxExistenceMap. So we need to it here.
func (*Handle) StartWorker ¶
func (h *Handle) StartWorker()
StartWorker starts the background collector worker inside
type MaxTidRecord ¶
type MaxTidRecord struct {
// contains filtered or unexported fields
}
MaxTidRecord is to record the max tid.