worker

package
v2.0.0-rc.2+incompatible Latest Latest
Warning

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

 Go to latest Published: Jun 22, 2016 License: BSD-3-Clause Imports: 47 Imported by: 0

Documentation

Overview

Package worker contains the framework, utility methods and core functions for long running actions. 'vtworker' binary will use these.

Index

Constants

This section is empty.

Variables

This section is empty.

Functions

func AddCommand 

func AddCommand(groupName string, c Command)

AddCommand registers a command and makes it available.

func CompareRows 

func CompareRows(fields []*querypb.Field, compareCount int, left, right []sqltypes.Value) (int, error)

CompareRows returns: -1 if left is smaller than right 0 if left and right are equal +1 if left is bigger than right TODO: This can panic if types for left and right don't match.

func FindChunks 

func FindChunks(ctx context.Context, wr *wrangler.Wrangler, ti *topo.TabletInfo, td *tabletmanagerdatapb.TableDefinition, minTableSizeForSplit uint64, sourceReaderCount int) ([]string, error)

FindChunks returns an array of chunks to use for splitting up a table into multiple data chunks. It only works for tables with a primary key (and the primary key first column is an integer type). The array will always look like: "", "value1", "value2", "" A non-split tablet will just return: "", ""

func FindHealthyRdonlyTablet 

func FindHealthyRdonlyTablet(ctx context.Context, wr *wrangler.Wrangler, cell, keyspace, shard string, minHealthyRdonlyTablets int) (*topodatapb.TabletAlias, error)

FindHealthyRdonlyTablet returns a random healthy RDONLY tablet. Since we don't want to use them all, we require at least minHealthyRdonlyTablets servers to be healthy. May block up to -wait_for_healthy_rdonly_tablets_timeout.

func FindWorkerTablet 

func FindWorkerTablet(ctx context.Context, wr *wrangler.Wrangler, cleaner *wrangler.Cleaner, cell, keyspace, shard string, minHealthyRdonlyTablets int) (*topodatapb.TabletAlias, error)

FindWorkerTablet will: - find a rdonly instance in the keyspace / shard - mark it as worker - tag it with our worker process

func PrintAllCommands 

func PrintAllCommands(logger logutil.Logger)

PrintAllCommands prints a help text for all registered commands to the given Logger.

func RowsEqual 

func RowsEqual(left, right []sqltypes.Value) int

RowsEqual returns the index of the first different fields, or -1 if both rows are the same

Types

type BlockWorker 

type BlockWorker struct {
	StatusWorker
	// contains filtered or unexported fields
}

BlockWorker will block infinitely until its context is cancelled.

func (*BlockWorker) Run 

func (bw *BlockWorker) Run(ctx context.Context) error

Run implements the Worker interface.

func (*BlockWorker) StatusAsHTML 

func (bw *BlockWorker) StatusAsHTML() template.HTML

StatusAsHTML implements the Worker interface.

func (*BlockWorker) StatusAsText 

func (bw *BlockWorker) StatusAsText() string

StatusAsText implements the Worker interface.

type Command 

type Command struct {
	Name        string
	Method      func(wi *Instance, wr *wrangler.Wrangler, subFlags *flag.FlagSet, args []string) (Worker, error)
	Interactive func(ctx context.Context, wi *Instance, wr *wrangler.Wrangler, w http.ResponseWriter, r *http.Request) (Worker, *template.Template, map[string]interface{}, error)
	Params      string
	Help        string // if help is empty, won't list the command
}

Command contains the detail of a command which can be run in vtworker. While "Method" is run from the command line or RPC, "Interactive" may contain special logic to parse a web form and return templated HTML output.

type DiffReport 

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

DiffReport has the stats for a diff job

func (*DiffReport) ComputeQPS 

func (dr *DiffReport) ComputeQPS()

ComputeQPS fills in processingQPS

func (*DiffReport) HasDifferences 

func (dr *DiffReport) HasDifferences() bool

HasDifferences returns true if the diff job recorded any difference

func (*DiffReport) String 

func (dr *DiffReport) String() string

type Instance 

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

Instance encapsulate the execution state of vtworker.

func NewInstance 

func NewInstance(ctx context.Context, ts topo.Server, cell string, commandDisplayInterval time.Duration) *Instance

NewInstance creates a new Instance.

func (*Instance) Cancel 

func (wi *Instance) Cancel() bool

Cancel calls the cancel function of the current vtworker job. It returns true, if a job was running. False otherwise. NOTE: Cancel won't reset the state as well. Use Reset() to do so.

func (*Instance) CreateWrangler 

func (wi *Instance) CreateWrangler(logger logutil.Logger) *wrangler.Wrangler

CreateWrangler creates a new wrangler using the instance specific configuration.

func (*Instance) InitInteractiveMode 

func (wi *Instance) InitInteractiveMode()

InitInteractiveMode installs webserver handlers for each known command.

func (*Instance) InitStatusHandling 

func (wi *Instance) InitStatusHandling()

InitStatusHandling installs webserver handlers for global actions like /status, /reset and /cancel.

func (*Instance) InstallSignalHandlers 

func (wi *Instance) InstallSignalHandlers()

InstallSignalHandlers installs signal handler which exit vtworker gracefully.

func (*Instance) Reset 

func (wi *Instance) Reset() error

Reset resets the state of a finished worker. It returns an error if the worker is still running.

func (*Instance) RunCommand 

func (wi *Instance) RunCommand(args []string, wr *wrangler.Wrangler, runFromCli bool) (Worker, chan struct{}, error)

RunCommand executes the vtworker command specified by "args". Use WaitForCommand() to block on the returned done channel. If wr is nil, the default wrangler will be used. If you pass a wr wrangler, note that a MemoryLogger will be added to its current logger. The returned worker and done channel may be nil if no worker was started e.g. in case of a "Reset".

func (*Instance) WaitForCommand 

func (wi *Instance) WaitForCommand(wrk Worker, done chan struct{}) error

WaitForCommand blocks until "done" is closed. In the meantime, it logs the status of "wrk".

type PanicWorker 

type PanicWorker struct {
	StatusWorker
	// contains filtered or unexported fields
}

PanicWorker will run panic() when executed. For internal tests only.

func (*PanicWorker) Run 

func (pw *PanicWorker) Run(ctx context.Context) error

Run implements the Worker interface.

func (*PanicWorker) StatusAsHTML 

func (pw *PanicWorker) StatusAsHTML() template.HTML

StatusAsHTML implements the Worker interface

func (*PanicWorker) StatusAsText 

func (pw *PanicWorker) StatusAsText() string

StatusAsText implements the Worker interface.

type PingWorker 

type PingWorker struct {
	StatusWorker
	// contains filtered or unexported fields
}

PingWorker will log a message with level CONSOLE.

func (*PingWorker) Run 

func (pw *PingWorker) Run(ctx context.Context) error

Run implements the Worker interface.

func (*PingWorker) StatusAsHTML 

func (pw *PingWorker) StatusAsHTML() template.HTML

StatusAsHTML implements the Worker interface

func (*PingWorker) StatusAsText 

func (pw *PingWorker) StatusAsText() string

StatusAsText implements the Worker interface.

type QueryResultReader 

type QueryResultReader struct {
	Output sqltypes.ResultStream
	Fields []*querypb.Field
	// contains filtered or unexported fields
}

QueryResultReader will stream rows towards the output channel.

func NewQueryResultReaderForTablet 

func NewQueryResultReaderForTablet(ctx context.Context, ts topo.Server, tabletAlias *topodatapb.TabletAlias, sql string) (*QueryResultReader, error)

NewQueryResultReaderForTablet creates a new QueryResultReader for the provided tablet / sql query

func TableScan 

func TableScan(ctx context.Context, log logutil.Logger, ts topo.Server, tabletAlias *topodatapb.TabletAlias, tableDefinition *tabletmanagerdatapb.TableDefinition) (*QueryResultReader, error)

TableScan returns a QueryResultReader that gets all the rows from a table, ordered by Primary Key. The returned columns are ordered with the Primary Key columns in front.

func TableScanByKeyRange 

func TableScanByKeyRange(ctx context.Context, log logutil.Logger, ts topo.Server, tabletAlias *topodatapb.TabletAlias, tableDefinition *tabletmanagerdatapb.TableDefinition, keyRange *topodatapb.KeyRange, keyspaceSchema *vindexes.KeyspaceSchema, shardingColumnName string, shardingColumnType topodatapb.KeyspaceIdType) (*QueryResultReader, error)

TableScanByKeyRange returns a QueryResultReader that gets all the rows from a table that match the supplied KeyRange, ordered by Primary Key. The returned columns are ordered with the Primary Key columns in front. If keyspaceSchema is passed in, we go into v3 mode, and we ask for all source data, and filter here. Otherwise we stick with v2 mode, where we can ask the source tablet to do the filtering.

func (*QueryResultReader) Close 

func (qrr *QueryResultReader) Close()

Close closes the connection to the tablet.

type RowDiffer 

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

RowDiffer will consume rows on both sides, and compare them. It assumes left and right are sorted by ascending primary key. it will record errors if extra rows exist on either side.

func NewRowDiffer 

func NewRowDiffer(left, right *QueryResultReader, tableDefinition *tabletmanagerdatapb.TableDefinition) (*RowDiffer, error)

NewRowDiffer returns a new RowDiffer

func (*RowDiffer) Go 

func (rd *RowDiffer) Go(log logutil.Logger) (dr DiffReport, err error)

Go runs the diff. If there is no error, it will drain both sides. If an error occurs, it will just return it and stop.

type RowReader 

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

RowReader returns individual rows from a QueryResultReader

func NewRowReader 

func NewRowReader(queryResultReader *QueryResultReader) *RowReader

NewRowReader returns a RowReader based on the QueryResultReader

func (*RowReader) Drain 

func (rr *RowReader) Drain() (int, error)

Drain will empty the RowReader and return how many rows we got

func (*RowReader) Fields 

func (rr *RowReader) Fields() []*querypb.Field

Fields returns the types for the rows

func (*RowReader) Next 

func (rr *RowReader) Next() ([]sqltypes.Value, error)

Next will return: (row, nil) for the next row (nil, nil) for EOF (nil, error) if an error occurred

type RowSplitter 

type RowSplitter struct {
	KeyResolver keyspaceIDResolver
	KeyRanges   []*topodatapb.KeyRange
}

RowSplitter is a helper class to split rows into multiple subsets targeted to different shards.

func NewRowSplitter 

func NewRowSplitter(shardInfos []*topo.ShardInfo, keyResolver keyspaceIDResolver) *RowSplitter

NewRowSplitter returns a new row splitter for the given shard distribution.

func (*RowSplitter) Send 

func (rs *RowSplitter) Send(fields []*querypb.Field, result [][][]sqltypes.Value, baseCmd string, insertChannels []chan string, abort <-chan struct{}) bool

Send will send the rows to the list of channels. Returns true if aborted.

func (*RowSplitter) Split 

func (rs *RowSplitter) Split(result [][][]sqltypes.Value, rows [][]sqltypes.Value) error

Split will split the rows into subset for each distribution

func (*RowSplitter) StartSplit 

func (rs *RowSplitter) StartSplit() [][][]sqltypes.Value

StartSplit starts a new split. Split can then be called multiple times.

type RowSubsetDiffer 

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

RowSubsetDiffer will consume rows on both sides, and compare them. It assumes superset and subset are sorted by ascending primary key. It will record errors in DiffReport.extraRowsRight if extra rows exist on the subset side, and DiffReport.extraRowsLeft will always be zero.

func NewRowSubsetDiffer 

func NewRowSubsetDiffer(superset, subset *QueryResultReader, pkFieldCount int) (*RowSubsetDiffer, error)

NewRowSubsetDiffer returns a new RowSubsetDiffer

func (*RowSubsetDiffer) Go 

func (rd *RowSubsetDiffer) Go(log logutil.Logger) (dr DiffReport, err error)

Go runs the diff. If there is no error, it will drain both sides. If an error occurs, it will just return it and stop.

type SplitCloneWorker 

type SplitCloneWorker struct {
	StatusWorker
	// contains filtered or unexported fields
}

SplitCloneWorker will clone the data within a keyspace from a source set of shards to a destination set of shards.

func (*SplitCloneWorker) Run 

func (scw *SplitCloneWorker) Run(ctx context.Context) error

Run implements the Worker interface

func (*SplitCloneWorker) StatusAsHTML 

func (scw *SplitCloneWorker) StatusAsHTML() template.HTML

StatusAsHTML implements the Worker interface

func (*SplitCloneWorker) StatusAsText 

func (scw *SplitCloneWorker) StatusAsText() string

StatusAsText implements the Worker interface

type SplitDiffWorker 

type SplitDiffWorker struct {
	StatusWorker
	// contains filtered or unexported fields
}

SplitDiffWorker executes a diff between a destination shard and its source shards in a shard split case.

func (*SplitDiffWorker) Run 

func (sdw *SplitDiffWorker) Run(ctx context.Context) error

Run is mostly a wrapper to run the cleanup at the end.

func (*SplitDiffWorker) StatusAsHTML 

func (sdw *SplitDiffWorker) StatusAsHTML() template.HTML

StatusAsHTML is part of the Worker interface

func (*SplitDiffWorker) StatusAsText 

func (sdw *SplitDiffWorker) StatusAsText() string

StatusAsText is part of the Worker interface

type StatusWorker 

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

StatusWorker is the base type for a worker which keeps a status. The status is protected by a mutex. StatusWorker also provides default implementations for StatusAsHTML and StatusAsText to make it easier on workers if they don't need to export more.

func NewStatusWorker 

func NewStatusWorker() StatusWorker

NewStatusWorker returns a StatusWorker in state WorkerStateNotStarted.

func (*StatusWorker) SetState 

func (w *StatusWorker) SetState(state StatusWorkerState)

SetState is a convenience function for workers.

func (*StatusWorker) State 

func (w *StatusWorker) State() StatusWorkerState

State is part of the Worker interface.

func (*StatusWorker) StatusAsHTML 

func (w *StatusWorker) StatusAsHTML() template.HTML

StatusAsHTML is part of the Worker interface.

func (*StatusWorker) StatusAsText 

func (w *StatusWorker) StatusAsText() string

StatusAsText is part of the Worker interface.

type StatusWorkerState 

type StatusWorkerState string

StatusWorkerState is the type for a StatusWorker's status 

const (
	// WorkerStateNotStarted is the initial state.
	WorkerStateNotStarted StatusWorkerState = "not started"
	// WorkerStateDone is set when the worker successfully finished.
	WorkerStateDone StatusWorkerState = "done"
	// WorkerStateError is set when the worker failed.
	WorkerStateError StatusWorkerState = "error"
	// WorkerStateInit is set when the worker does initialize its state.
	WorkerStateInit StatusWorkerState = "initializing"
	// WorkerStateFindTargets is set when the worker searches healthy RDONLY tablets.
	WorkerStateFindTargets StatusWorkerState = "finding target instances"
	// WorkerStateSyncReplication is set when the worker ensures that source and
	// destination tablets are at the same GTID during the diff.
	WorkerStateSyncReplication StatusWorkerState = "synchronizing replication"
	// WorkerStateCopy is set when the worker copies the data.
	WorkerStateCopy StatusWorkerState = "copying the data"
	// WorkerStateDiff is set when the worker compares the data.
	WorkerStateDiff StatusWorkerState = "running the diff"
	// WorkerStateCleanUp is set when the worker reverses the initialization e.g.
	// the type of a taken out RDONLY tablet is changed back from "worker" to "spare".
	WorkerStateCleanUp StatusWorkerState = "cleaning up"
)

func (StatusWorkerState) String 

func (state StatusWorkerState) String() string

type VerticalSplitCloneWorker 

type VerticalSplitCloneWorker struct {
	StatusWorker
	// contains filtered or unexported fields
}

VerticalSplitCloneWorker will clone the data from a source keyspace/shard to a destination keyspace/shard.

func (*VerticalSplitCloneWorker) Run 

func (vscw *VerticalSplitCloneWorker) Run(ctx context.Context) error

Run implements the Worker interface

func (*VerticalSplitCloneWorker) StatusAsHTML 

func (vscw *VerticalSplitCloneWorker) StatusAsHTML() template.HTML

StatusAsHTML implements the Worker interface

func (*VerticalSplitCloneWorker) StatusAsText 

func (vscw *VerticalSplitCloneWorker) StatusAsText() string

StatusAsText implements the Worker interface

type VerticalSplitDiffWorker 

type VerticalSplitDiffWorker struct {
	StatusWorker
	// contains filtered or unexported fields
}

VerticalSplitDiffWorker executes a diff between a destination shard and its source shards in a shard split case.

func (*VerticalSplitDiffWorker) Run 

func (vsdw *VerticalSplitDiffWorker) Run(ctx context.Context) error

Run is mostly a wrapper to run the cleanup at the end.

func (*VerticalSplitDiffWorker) StatusAsHTML 

func (vsdw *VerticalSplitDiffWorker) StatusAsHTML() template.HTML

StatusAsHTML is part of the Worker interface.

func (*VerticalSplitDiffWorker) StatusAsText 

func (vsdw *VerticalSplitDiffWorker) StatusAsText() string

StatusAsText is part of the Worker interface.

type Worker 

type Worker interface {
	// State returns the current state using the internal representation.
	State() StatusWorkerState

	// StatusAsHTML returns the current worker status in HTML.
	StatusAsHTML() template.HTML

	// StatusAsText returns the current worker status in plain text.
	StatusAsText() string

	// Run is the main entry point for the worker. It will be
	// called in a go routine.  When the passed in context is
	// cancelled, Run should exit as soon as possible.
	Run(context.Context) error
}

Worker is the base interface for all long running workers.

func NewBlockWorker 

func NewBlockWorker(wr *wrangler.Wrangler) (Worker, error)

NewBlockWorker returns a new BlockWorker object.

func NewPanicWorker 

func NewPanicWorker(wr *wrangler.Wrangler) (Worker, error)

NewPanicWorker returns a new PanicWorker object.

func NewPingWorker 

func NewPingWorker(wr *wrangler.Wrangler, message string) (Worker, error)

NewPingWorker returns a new PingWorker object.

func NewSplitCloneWorker 

func NewSplitCloneWorker(wr *wrangler.Wrangler, cell, keyspace, shard string, excludeTables []string, strategyStr string, sourceReaderCount, destinationPackCount int, minTableSizeForSplit uint64, destinationWriterCount, minHealthyRdonlyTablets int, maxTPS int64) (Worker, error)

NewSplitCloneWorker returns a new SplitCloneWorker object.

func NewSplitDiffWorker 

func NewSplitDiffWorker(wr *wrangler.Wrangler, cell, keyspace, shard string, sourceUID uint32, excludeTables []string, minHealthyRdonlyTablets int) Worker

NewSplitDiffWorker returns a new SplitDiffWorker object.

func NewVerticalSplitCloneWorker 

func NewVerticalSplitCloneWorker(wr *wrangler.Wrangler, cell, destinationKeyspace, destinationShard string, tables []string, strategyStr string, sourceReaderCount, destinationPackCount int, minTableSizeForSplit uint64, destinationWriterCount, minHealthyRdonlyTablets int, maxTPS int64) (Worker, error)

NewVerticalSplitCloneWorker returns a new VerticalSplitCloneWorker object.

func NewVerticalSplitDiffWorker 

func NewVerticalSplitDiffWorker(wr *wrangler.Wrangler, cell, keyspace, shard string, minHealthyRdonlyTablets int) Worker

NewVerticalSplitDiffWorker returns a new VerticalSplitDiffWorker object.

Source Files

View all Source files

Directories

Path Synopsis
Package fakevtworkerclient contains a fake for the vtworkerclient interface.
 Package fakevtworkerclient contains a fake for the vtworkerclient interface.
Package grpcvtworkerclient contains the gRPC version of the vtworker client protocol.
 Package grpcvtworkerclient contains the gRPC version of the vtworker client protocol.
Package grpcvtworkerserver contains the gRPC implementation of the server side of the remote execution of vtworker commands.
 Package grpcvtworkerserver contains the gRPC implementation of the server side of the remote execution of vtworker commands.
Package vtworkerclient contains the generic client side of the remote vtworker protocol.
 Package vtworkerclient contains the generic client side of the remote vtworker protocol.
Package vtworkerclienttest contains the testsuite against which each RPC implementation of the vtworkerclient interface must be tested.
 Package vtworkerclienttest contains the testsuite against which each RPC implementation of the vtworkerclient interface must be tested.

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.