discovery

package
v0.21.1 Latest Latest
Warning

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

 Go to latest Published: Dec 3, 2024 License: Apache-2.0 Imports: 44 Imported by: 28

Documentation

Overview

Package discovery provides a way to discover all tablets e.g. within a specific shard and monitor their current health.

Use the HealthCheck object to query for tablets and their health.

For an example how to use the HealthCheck object, see vtgate/tabletgateway.go

Tablets have to be manually added to the HealthCheck using AddTablet(). Alternatively, use a Watcher implementation which will constantly watch a source (e.g. the topology) and add and remove tablets as they are added or removed from the source. For a Watcher example have a look at NewTopologyWatcher().

Internally, the HealthCheck module is connected to each tablet and has a streaming RPC (StreamHealth) open to receive periodic health infos.

Index

Constants

View Source 
const (
	DefaultHealthCheckRetryDelay = 5 * time.Second
	DefaultHealthCheckTimeout    = 1 * time.Minute
)

See the documentation for NewHealthCheck below for an explanation of these parameters.

Variables

View Source 
var (

	// TabletURLTemplateString is a flag to generate URLs for the tablets that vtgate discovers.
	TabletURLTemplateString = "http://{{.GetTabletHostPort}}"

	// AllowedTabletTypes is the list of allowed tablet types. e.g. {PRIMARY, REPLICA}.
	AllowedTabletTypes []topodata.TabletType

	// KeyspacesToWatch - if provided this specifies which keyspaces should be
	// visible to the healthcheck. By default the healthcheck will watch all keyspaces.
	KeyspacesToWatch []string

	// HealthCheckCacheTemplate uses healthCheckTemplate with the `HealthCheck Tablet - Cache` title to create the
	// HTML code required to render the cache of the HealthCheck.
	HealthCheckCacheTemplate = fmt.Sprintf(healthCheckTemplate, "HealthCheck - Cache")

	// HealthCheckHealthyTemplate uses healthCheckTemplate with the `HealthCheck Tablet - Healthy Tablets` title to
	// create the HTML code required to render the list of healthy tablets from the HealthCheck.
	HealthCheckHealthyTemplate = fmt.Sprintf(healthCheckTemplate, "HealthCheck - Healthy Tablets")
)

Functions

func BuildTabletTypesString added in v0.20.0 

func BuildTabletTypesString(tabletTypes []topodatapb.TabletType, tabletSelectionPreference tabletmanagerdatapb.TabletSelectionPreference) string

BuildTabletTypesString is a helper to build a serialized string representation of the tablet type(s) and optional in order clause for later use with the TabletPicker.

func FilteringKeyspaces added in v0.10.0 

func FilteringKeyspaces() bool

FilteringKeyspaces returns true if any keyspaces have been configured to be filtered.

func GetHighReplicationLagMinServing added in v0.11.0 

func GetHighReplicationLagMinServing() time.Duration

GetHighReplicationLagMinServing getter for use by debugenv

func GetLowReplicationLag added in v0.11.0 

func GetLowReplicationLag() time.Duration

GetLowReplicationLag getter for use by debugenv

func GetMinNumTablets added in v0.11.0 

func GetMinNumTablets() int

GetMinNumTablets getter for use by debugenv

func GetTabletPickerRetryDelay added in v0.8.0 

func GetTabletPickerRetryDelay() time.Duration

GetTabletPickerRetryDelay synchronizes changes to tabletPickerRetryDelay. Used in tests only at the moment

func IsReplicationLagHigh 

func IsReplicationLagHigh(tabletHealth *TabletHealth) bool

IsReplicationLagHigh verifies that the given TabletHealth refers to a tablet with high replication lag, i.e. higher than the configured discovery_low_replication_lag flag.

func IsReplicationLagVeryHigh 

func IsReplicationLagVeryHigh(tabletHealth *TabletHealth) bool

IsReplicationLagVeryHigh verifies that the given TabletHealth refers to a tablet with very high replication lag, i.e. higher than the configured discovery_high_replication_lag_minimum_serving flag.

func ParseTabletTypesAndOrder added in v0.14.0 

func ParseTabletTypesAndOrder(tabletTypesStr string) ([]topodatapb.TabletType, bool, error)

func ParseTabletURLTemplateFromFlag 

func ParseTabletURLTemplateFromFlag()

ParseTabletURLTemplateFromFlag loads or reloads the URL template.

func SetHighReplicationLagMinServing added in v0.11.0 

func SetHighReplicationLagMinServing(lag time.Duration)

SetHighReplicationLagMinServing setter for use by debugenv

func SetLowReplicationLag added in v0.11.0 

func SetLowReplicationLag(lag time.Duration)

SetLowReplicationLag setter for use by debugenv

func SetMinNumTablets added in v0.11.0 

func SetMinNumTablets(numTablets int)

SetMinNumTablets setter for use by debugenv

func SetTabletPickerRetryDelay added in v0.8.0 

func SetTabletPickerRetryDelay(delay time.Duration)

SetTabletPickerRetryDelay synchronizes reads for tabletPickerRetryDelay. Used in tests only at the moment

func TabletToMapKey 

func TabletToMapKey(tablet *topodata.Tablet) string

TabletToMapKey creates a key to the map from tablet's host and ports. It should only be used in discovery and related module.

Types

type FakeHealthCheck 

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

FakeHealthCheck implements discovery.HealthCheck.

func NewFakeHealthCheck 

func NewFakeHealthCheck(ch chan *TabletHealth) *FakeHealthCheck

NewFakeHealthCheck returns the fake healthcheck object.

func (*FakeHealthCheck) AddFakeTablet 

func (fhc *FakeHealthCheck) AddFakeTablet(cell, host string, port int32, keyspace, shard string, tabletType topodatapb.TabletType, serving bool, reparentTS int64, err error, connFactory func(*topodatapb.Tablet) queryservice.QueryService) queryservice.QueryService

AddFakeTablet inserts a fake entry into FakeHealthCheck. The Tablet can be talked to using the provided connection. The Listener is called, as if AddTablet had been called. For flexibility the connection is created via a connFactory callback

func (*FakeHealthCheck) AddTablet 

func (fhc *FakeHealthCheck) AddTablet(tablet *topodatapb.Tablet)

AddTablet adds the tablet.

func (*FakeHealthCheck) AddTestTablet 

func (fhc *FakeHealthCheck) AddTestTablet(cell, host string, port int32, keyspace, shard string, tabletType topodatapb.TabletType, serving bool, reparentTS int64, err error) *sandboxconn.SandboxConn

AddTestTablet adds a fake tablet for tests using the SandboxConn and returns the fake connection

func (*FakeHealthCheck) Broadcast added in v0.13.0 

func (fhc *FakeHealthCheck) Broadcast(tablet *topodatapb.Tablet)

Broadcast broadcasts the healthcheck for the given tablet

func (*FakeHealthCheck) BroadcastAll added in v0.18.0 

func (fhc *FakeHealthCheck) BroadcastAll()

BroadcastAll broadcasts all the tablets' healthchecks

func (*FakeHealthCheck) CacheStatus 

func (fhc *FakeHealthCheck) CacheStatus() TabletsCacheStatusList

CacheStatus returns the status for each tablet

func (*FakeHealthCheck) CacheStatusMap added in v0.14.0 

func (fhc *FakeHealthCheck) CacheStatusMap() map[string]*TabletsCacheStatus

CacheStatusMap returns a map of the health check cache.

func (*FakeHealthCheck) Close 

func (fhc *FakeHealthCheck) Close() error

Close is not implemented.

func (*FakeHealthCheck) GetAllTablets 

func (fhc *FakeHealthCheck) GetAllTablets() map[string]*topodatapb.Tablet

GetAllTablets returns all the tablets we have.

func (*FakeHealthCheck) GetHealthyTabletStats 

func (fhc *FakeHealthCheck) GetHealthyTabletStats(target *querypb.Target) []*TabletHealth

GetHealthyTabletStats returns only the healthy tablets - Serving true and LastError is not nil

func (*FakeHealthCheck) GetLoadTabletsTrigger added in v0.17.0 

func (fhc *FakeHealthCheck) GetLoadTabletsTrigger() chan struct{}

GetLoadTabletsTrigger is not implemented.

func (*FakeHealthCheck) GetPrimaryTablet added in v0.13.0 

func (fhc *FakeHealthCheck) GetPrimaryTablet() *topodatapb.Tablet

GetPrimaryTablet gets the primary tablet from the tablets that healthcheck has seen so far

func (*FakeHealthCheck) GetTabletHealth added in v0.14.0 

func (fhc *FakeHealthCheck) GetTabletHealth(kst KeyspaceShardTabletType, alias *topodatapb.TabletAlias) (*TabletHealth, error)

GetTabletHealth results the TabletHealth of the tablet that matches the given alias

func (*FakeHealthCheck) GetTabletHealthByAlias added in v0.14.0 

func (fhc *FakeHealthCheck) GetTabletHealthByAlias(alias *topodatapb.TabletAlias) (*TabletHealth, error)

GetTabletHealthByAlias results the TabletHealth of the tablet that matches the given alias

func (*FakeHealthCheck) HealthyStatus added in v0.19.0 

func (fhc *FakeHealthCheck) HealthyStatus() TabletsCacheStatusList

HealthyStatus returns the status for each healthy tablet

func (*FakeHealthCheck) RegisterStats 

func (fhc *FakeHealthCheck) RegisterStats()

RegisterStats is not implemented.

func (*FakeHealthCheck) RemoveTablet 

func (fhc *FakeHealthCheck) RemoveTablet(tablet *topodatapb.Tablet)

RemoveTablet removes the tablet.

func (*FakeHealthCheck) ReplaceTablet 

func (fhc *FakeHealthCheck) ReplaceTablet(old, new *topodatapb.Tablet)

ReplaceTablet removes the old tablet and adds the new.

func (*FakeHealthCheck) Reset 

func (fhc *FakeHealthCheck) Reset()

Reset cleans up the internal state.

func (*FakeHealthCheck) SetPrimaryTimestamp added in v0.19.8 

func (fhc *FakeHealthCheck) SetPrimaryTimestamp(tablet *topodatapb.Tablet, timestamp int64)

SetPrimaryTimestamp sets the primary timestamp for the given tablet

func (*FakeHealthCheck) SetServing added in v0.13.0 

func (fhc *FakeHealthCheck) SetServing(tablet *topodatapb.Tablet, serving bool)

SetServing sets the serving variable for the given tablet

func (*FakeHealthCheck) SetTabletType added in v0.13.0 

func (fhc *FakeHealthCheck) SetTabletType(tablet *topodatapb.Tablet, tabletType topodatapb.TabletType)

SetTabletType sets the tablet type for the given tablet

func (*FakeHealthCheck) Subscribe 

func (fhc *FakeHealthCheck) Subscribe() chan *TabletHealth

Subscribe returns the channel in the struct. Subscribe should only be called in one place for this fake health check

func (*FakeHealthCheck) TabletConnection 

func (fhc *FakeHealthCheck) TabletConnection(ctx context.Context, alias *topodatapb.TabletAlias, target *querypb.Target) (queryservice.QueryService, error)

TabletConnection returns the TabletConn of the given tablet.

func (*FakeHealthCheck) Unsubscribe 

func (fhc *FakeHealthCheck) Unsubscribe(c chan *TabletHealth)

Unsubscribe is not implemented.

func (*FakeHealthCheck) UpdateHealth added in v0.14.0 

func (fhc *FakeHealthCheck) UpdateHealth(th *TabletHealth)

UpdateHealth adds a TabletHealth for the tablet defined in th.

func (*FakeHealthCheck) WaitForAllServingTablets 

func (fhc *FakeHealthCheck) WaitForAllServingTablets(ctx context.Context, targets []*querypb.Target) error

WaitForAllServingTablets is not implemented.

type FilterByKeyspace 

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

FilterByKeyspace is a filter that filters tablets by keyspace.

func NewFilterByKeyspace 

func NewFilterByKeyspace(selectedKeyspaces []string) *FilterByKeyspace

NewFilterByKeyspace creates a new FilterByKeyspace. Each filter is a keyspace entry. All tablets that match a keyspace will be forwarded to the TopologyWatcher's consumer.

func (*FilterByKeyspace) IsIncluded 

func (fbk *FilterByKeyspace) IsIncluded(tablet *topodata.Tablet) bool

IsIncluded returns true if the tablet's keyspace matches what we have.

type FilterByShard 

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

FilterByShard is a filter that filters tablets by keyspace/shard.

func NewFilterByShard 

func NewFilterByShard(filters []string) (*FilterByShard, error)

NewFilterByShard creates a new FilterByShard for use by a TopologyWatcher. Each filter is a keyspace|shard entry, where shard can either be a shard name, or a keyrange. All tablets that match at least one keyspace|shard tuple will be forwarded by the TopologyWatcher to its consumer.

func (*FilterByShard) IsIncluded 

func (fbs *FilterByShard) IsIncluded(tablet *topodata.Tablet) bool

IsIncluded returns true iff the tablet's keyspace and shard match what we have.

type FilterByTabletTags added in v0.20.0 

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

FilterByTabletTags is a filter that filters tablets by tablet tag key/values.

func NewFilterByTabletTags added in v0.20.0 

func NewFilterByTabletTags(tabletTags map[string]string) *FilterByTabletTags

NewFilterByTabletTags creates a new FilterByTabletTags. All tablets that match all tablet tags will be forwarded to the TopologyWatcher's consumer.

func (*FilterByTabletTags) IsIncluded added in v0.20.0 

func (fbtg *FilterByTabletTags) IsIncluded(tablet *topodata.Tablet) bool

IsIncluded returns true if the tablet's tags match what we expect.

type HealthCheck 

type HealthCheck interface {
	// AddTablet adds the tablet.
	AddTablet(tablet *topodata.Tablet)

	// RemoveTablet removes the tablet.
	RemoveTablet(tablet *topodata.Tablet)

	// ReplaceTablet does an AddTablet and RemoveTablet in one call, effectively replacing the old tablet with the new.
	ReplaceTablet(old, new *topodata.Tablet)

	// CacheStatus returns a displayable version of the health check cache.
	CacheStatus() TabletsCacheStatusList

	// HealthyStatus returns a displayable version of the health check healthy list.
	HealthyStatus() TabletsCacheStatusList

	// CacheStatusMap returns a map of the health check cache.
	CacheStatusMap() map[string]*TabletsCacheStatus

	// Close stops the healthcheck.
	Close() error

	// WaitForAllServingTablets waits for at least one healthy serving tablet in
	// each given target before returning.
	// It will return ctx.Err() if the context is canceled.
	// It will return an error if it can't read the necessary topology records.
	WaitForAllServingTablets(ctx context.Context, targets []*query.Target) error

	// TabletConnection returns the TabletConn of the given tablet.
	TabletConnection(ctx context.Context, alias *topodata.TabletAlias, target *query.Target) (queryservice.QueryService, error)

	// RegisterStats registers the connection counts stats
	RegisterStats()

	// GetHealthyTabletStats returns only the healthy tablets.
	// The returned array is owned by the caller.
	// For TabletType_PRIMARY, this will only return at most one entry,
	// the most recent tablet of type primary.
	// This returns a copy of the data so that callers can access without
	// synchronization
	GetHealthyTabletStats(target *query.Target) []*TabletHealth

	// GetTabletHealth results the TabletHealth of the tablet that matches the given alias
	GetTabletHealth(kst KeyspaceShardTabletType, alias *topodata.TabletAlias) (*TabletHealth, error)

	// GetTabletHealthByAlias results the TabletHealth of the tablet that matches the given alias
	GetTabletHealthByAlias(alias *topodata.TabletAlias) (*TabletHealth, error)

	// Subscribe adds a listener. Used by vtgate buffer to learn about primary changes.
	Subscribe() chan *TabletHealth

	// Unsubscribe removes a listener.
	Unsubscribe(c chan *TabletHealth)

	// GetLoadTabletsTrigger returns a channel that is used to inform when to load tablets.
	GetLoadTabletsTrigger() chan struct{}
}

HealthCheck declares what the TabletGateway needs from the HealthCheck

type HealthCheckImpl 

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

HealthCheckImpl performs health checking and stores the results. The goal of this object is to maintain a StreamHealth RPC to a lot of tablets. Tablets are added / removed by calling the AddTablet / RemoveTablet methods (other discovery module objects can for instance watch the topology and call these). It contains a map of tabletHealthCheck objects by Alias. Each tabletHealthCheck object stores the health information for one tablet. A checkConn goroutine is spawned for each tabletHealthCheck, which is responsible for keeping that tabletHealthCheck up-to-date. If checkConn terminates for any reason, then the corresponding tabletHealthCheck object is removed from the map. When a tabletHealthCheck gets removed from the map, its cancelFunc gets called, which ensures that the associated checkConn goroutine eventually terminates.

func NewHealthCheck 

func NewHealthCheck(ctx context.Context, retryDelay, healthCheckTimeout time.Duration, topoServer *topo.Server, localCell, cellsToWatch string, filters TabletFilter) *HealthCheckImpl

NewHealthCheck creates a new HealthCheck object. Parameters: retryDelay. 

The duration to wait before retrying to connect (e.g. after a failed connection
attempt).

healthCheckTimeout. 

The duration for which we consider a health check response to be 'fresh'. If we don't get
a health check response from a tablet for more than this duration, we consider the tablet
not healthy.

topoServer. 

The topology server that this healthcheck object can use to retrieve cell or tablet information

localCell. 

The localCell for this healthcheck

cellsToWatch. 

Is a list of cells to watch for tablets.

filters. 

Is one or more filters to apply when determining what tablets we want to stream healthchecks from.

func (*HealthCheckImpl) AddTablet 

func (hc *HealthCheckImpl) AddTablet(tablet *topodata.Tablet)

AddTablet adds the tablet, and starts health check. It does not block on making connection. name is an optional tag for the tablet, e.g. an alternative address.

func (*HealthCheckImpl) CacheStatus 

func (hc *HealthCheckImpl) CacheStatus() TabletsCacheStatusList

CacheStatus returns a displayable version of the cache.

func (*HealthCheckImpl) CacheStatusMap added in v0.14.0 

func (hc *HealthCheckImpl) CacheStatusMap() map[string]*TabletsCacheStatus

func (*HealthCheckImpl) Close 

func (hc *HealthCheckImpl) Close() error

Close stops the healthcheck.

func (*HealthCheckImpl) GetHealthyTabletStats 

func (hc *HealthCheckImpl) GetHealthyTabletStats(target *query.Target) []*TabletHealth

GetHealthyTabletStats returns only the healthy tablets. The returned array is owned by the caller. For TabletType_PRIMARY, this will only return at most one entry, the most recent tablet of type primary. This returns a copy of the data so that callers can access without synchronization

func (*HealthCheckImpl) GetLoadTabletsTrigger added in v0.17.0 

func (hc *HealthCheckImpl) GetLoadTabletsTrigger() chan struct{}

GetLoadTabletsTrigger returns a channel that is used to inform when to load tablets.

func (*HealthCheckImpl) GetTabletHealth added in v0.14.0 

func (hc *HealthCheckImpl) GetTabletHealth(kst KeyspaceShardTabletType, alias *topodata.TabletAlias) (*TabletHealth, error)

GetTabletHealth results the TabletHealth of the tablet that matches the given alias and KeyspaceShardTabletType The data is retrieved from the healthData map.

func (*HealthCheckImpl) GetTabletHealthByAlias added in v0.14.0 

func (hc *HealthCheckImpl) GetTabletHealthByAlias(alias *topodata.TabletAlias) (*TabletHealth, error)

GetTabletHealthByAlias results the TabletHealth of the tablet that matches the given alias

func (*HealthCheckImpl) GetTabletStats added in v0.15.0 

func (hc *HealthCheckImpl) GetTabletStats(target *query.Target) []*TabletHealth

GetTabletStats returns all tablets for the given target. The returned array is owned by the caller. For TabletType_PRIMARY, this will only return at most one entry, the most recent tablet of type primary.

func (*HealthCheckImpl) HealthyStatus added in v0.19.0 

func (hc *HealthCheckImpl) HealthyStatus() TabletsCacheStatusList

HealthyStatus returns a displayable version of the cache.

func (*HealthCheckImpl) HealthyStatusMap added in v0.19.0 

func (hc *HealthCheckImpl) HealthyStatusMap() map[string]*TabletsCacheStatus

func (*HealthCheckImpl) RegisterStats 

func (hc *HealthCheckImpl) RegisterStats()

RegisterStats registers the connection counts stats

func (*HealthCheckImpl) RemoveTablet 

func (hc *HealthCheckImpl) RemoveTablet(tablet *topodata.Tablet)

RemoveTablet removes the tablet, and stops the health check. It does not block.

func (*HealthCheckImpl) ReplaceTablet 

func (hc *HealthCheckImpl) ReplaceTablet(old, new *topodata.Tablet)

ReplaceTablet removes the old tablet and adds the new tablet.

func (*HealthCheckImpl) ServeHTTP 

func (hc *HealthCheckImpl) ServeHTTP(w http.ResponseWriter, _ *http.Request)

ServeHTTP is part of the http.Handler interface. It renders the current state of the discovery gateway tablet cache into json.

func (*HealthCheckImpl) Subscribe 

func (hc *HealthCheckImpl) Subscribe() chan *TabletHealth

Subscribe adds a listener. Used by vtgate buffer to learn about primary changes.

func (*HealthCheckImpl) TabletConnection 

func (hc *HealthCheckImpl) TabletConnection(ctx context.Context, alias *topodata.TabletAlias, target *query.Target) (queryservice.QueryService, error)

TabletConnection returns the Connection to a given tablet.

func (*HealthCheckImpl) Unsubscribe 

func (hc *HealthCheckImpl) Unsubscribe(c chan *TabletHealth)

Unsubscribe removes a listener.

func (*HealthCheckImpl) WaitForAllServingTablets 

func (hc *HealthCheckImpl) WaitForAllServingTablets(ctx context.Context, targets []*query.Target) error

WaitForAllServingTablets waits for at least one healthy serving tablet in each given target before returning. It will return ctx.Err() if the context is canceled. It will return an error if it can't read the necessary topology records.

func (*HealthCheckImpl) WaitForTablets 

func (hc *HealthCheckImpl) WaitForTablets(ctx context.Context, keyspace, shard string, tabletType topodata.TabletType) error

WaitForTablets waits for at least one tablet in the given keyspace / shard / tablet type before returning. The tablets do not have to be healthy. It will return ctx.Err() if the context is canceled.

type KeyspaceEvent added in v0.12.0 

type KeyspaceEvent struct {
	// Cell is the cell where the keyspace lives
	Cell string

	// Keyspace is the name of the keyspace which was (partially) unavailable and is now fully healthy
	Keyspace string

	// Shards is a list of all the shards in the keyspace, including their state after the event is resolved
	Shards []ShardEvent

	// MoveTablesState records the current state of an ongoing MoveTables workflow
	MoveTablesState MoveTablesState
}

KeyspaceEvent is yielded to all watchers when an availability event for a keyspace has been resolved

type KeyspaceEventWatcher added in v0.12.0 

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

KeyspaceEventWatcher is an auxiliary watcher that watches all availability incidents for all keyspaces in a Vitess cell and notifies listeners when the events have been resolved. Right now this is capable of detecting the end of failovers, both planned and unplanned, and the end of resharding operations.

The KeyspaceEventWatcher works by consolidating TabletHealth events from a HealthCheck stream, which is a peer-to-peer check between nodes via GRPC, with events from a Topology Server, which are global to the cluster and stored in an external system like etcd.

func NewKeyspaceEventWatcher added in v0.12.0 

func NewKeyspaceEventWatcher(ctx context.Context, topoServer srvtopo.Server, hc HealthCheck, localCell string) *KeyspaceEventWatcher

NewKeyspaceEventWatcher returns a new watcher for all keyspace events in the given cell. It requires access to a topology server, and an existing HealthCheck implementation which will be used to detect unhealthy nodes.

func (*KeyspaceEventWatcher) GetServingKeyspaces added in v0.18.0 

func (kew *KeyspaceEventWatcher) GetServingKeyspaces() []string

GetServingKeyspaces gets the serving keyspaces from the keyspace event watcher.

func (*KeyspaceEventWatcher) MarkShardNotServing added in v0.19.8 

func (kew *KeyspaceEventWatcher) MarkShardNotServing(ctx context.Context, keyspace string, shard string, isReparentErr bool) bool

MarkShardNotServing marks the given shard not serving. We use this when we start buffering for a given shard. This helps coordinate between the sharding logic and the keyspace event watcher. We take in a boolean as well to tell us whether this error is because a reparent is ongoing. If it is, we also mark the shard to wait for a reparent. The return argument is whether the shard was found and marked not serving successfully or not.

func (*KeyspaceEventWatcher) ShouldStartBufferingForTarget added in v0.21.0 

func (kew *KeyspaceEventWatcher) ShouldStartBufferingForTarget(ctx context.Context, target *querypb.Target) (*topodatapb.TabletAlias, bool)

ShouldStartBufferingForTarget checks if we should be starting buffering for the given target. We check the following things before we start buffering -

  1. The shard must have a primary.
  2. The primary must be non-serving.
  3. The keyspace must be marked inconsistent.

This buffering is meant to kick in during a Planned Reparent Shard operation. As part of that operation the old primary will become non-serving. At that point this code should return true to start buffering requests. Just as the PRS operation completes, a new primary will be elected, and it will send its own healthcheck stating that it is serving. We should buffer requests until that point.

There are use cases where people do not run with a Primary server at all, so we must verify that we only start buffering when a primary was present, and it went not serving. The shard state keeps track of the current primary and the last externally reparented time, which we can use to determine that there was a serving primary which now became non serving. This is only possible in a DemotePrimary RPC which are only called from ERS and PRS. So buffering will stop when these operations succeed. We also return the tablet alias of the primary if it is serving.

func (*KeyspaceEventWatcher) Subscribe added in v0.12.0 

func (kew *KeyspaceEventWatcher) Subscribe() chan *KeyspaceEvent

Subscribe returns a channel that will receive any KeyspaceEvents for all keyspaces in the current cell.

func (*KeyspaceEventWatcher) TargetIsBeingResharded added in v0.12.0 

func (kew *KeyspaceEventWatcher) TargetIsBeingResharded(ctx context.Context, target *querypb.Target) bool

TargetIsBeingResharded checks if the reason why the given target is not accessible right now is because the keyspace where it resides is (potentially) undergoing a resharding operation. This is not a fully accurate heuristic, but it's good enough that we'd want to buffer the request for the given target under the assumption that the reason why it cannot be completed right now is transitory.

func (*KeyspaceEventWatcher) Unsubscribe added in v0.12.0 

func (kew *KeyspaceEventWatcher) Unsubscribe(c chan *KeyspaceEvent)

Unsubscribe removes a listener previously returned from Subscribe

func (*KeyspaceEventWatcher) WaitForConsistentKeyspaces added in v0.21.0 

func (kew *KeyspaceEventWatcher) WaitForConsistentKeyspaces(ctx context.Context, ksList []string) error

WaitForConsistentKeyspaces waits for the given set of keyspaces to be marked consistent.

type KeyspaceShardTabletType added in v0.14.0 

type KeyspaceShardTabletType string

func KeyFromTarget added in v0.14.0 

func KeyFromTarget(target *query.Target) KeyspaceShardTabletType

KeyFromTarget includes cell which we ignore here because tabletStatsCache is intended to be per-cell

type MoveTablesState added in v0.18.0 

type MoveTablesState struct {
	Typ   MoveTablesType
	State MoveTablesStatus
}

func (MoveTablesState) String added in v0.20.0 

func (mts MoveTablesState) String() string

type MoveTablesStatus added in v0.18.0 

type MoveTablesStatus int
const (
	MoveTablesUnknown MoveTablesStatus = iota
	// MoveTablesSwitching is set when the write traffic is the middle of being switched from
	// the source to the target.
	MoveTablesSwitching
	// MoveTablesSwitched is set when write traffic has been completely switched to the target.
	MoveTablesSwitched
)

type MoveTablesType added in v0.18.0 

type MoveTablesType int
const (
	MoveTablesNone MoveTablesType = iota
	MoveTablesRegular
	MoveTablesShardByShard
)

type ShardEvent added in v0.12.0 

type ShardEvent struct {
	Tablet  *topodatapb.TabletAlias
	Target  *querypb.Target
	Serving bool
}

type TabletFilter 

type TabletFilter interface {
	// IsIncluded returns whether tablet is included in this filter
	IsIncluded(tablet *topodata.Tablet) bool
}

TabletFilter is an interface that can be given to a TopologyWatcher to be applied as an additional filter on the list of tablets returned by its getTablets function.

type TabletFilters 

type TabletFilters []TabletFilter

TabletFilters contains filters for tablets.

func NewVTGateHealthCheckFilters added in v0.19.7 

func NewVTGateHealthCheckFilters() (filters TabletFilters, err error)

NewVTGateHealthCheckFilters returns healthcheck filters for vtgate.

func (TabletFilters) IsIncluded added in v0.19.7 

func (tf TabletFilters) IsIncluded(tablet *topodata.Tablet) bool

IsIncluded returns true if a tablet passes all filters.

type TabletHealth 

type TabletHealth struct {
	Conn                 queryservice.QueryService
	Tablet               *topodata.Tablet
	Target               *query.Target
	Stats                *query.RealtimeStats
	PrimaryTermStartTime int64
	LastError            error
	Serving              bool
}

TabletHealth represents simple tablet health data that is returned to users of healthcheck. No synchronization is required because we always return a copy.

func FilterStatsByReplicationLag 

func FilterStatsByReplicationLag(tabletHealthList []*TabletHealth) []*TabletHealth

FilterStatsByReplicationLag filters the list of TabletHealth by TabletHealth.Stats.ReplicationLagSeconds. Note that TabletHealth that is non-serving or has error is ignored.

The simplified logic: - Return tablets that have lag <= lowReplicationLag. - Make sure we return at least minNumTablets tablets, if there are enough one with lag <= highReplicationLagMinServing. For example, with the default of 30s / 2h / 2, this means: - lags of (5s, 10s, 15s, 120s) return the first three - lags of (30m, 35m, 40m, 45m) return the first two - lags of (2h, 3h, 4h, 5h) return the first one

The legacy algorithm (default for now): - Return the list if there is 0 or 1 tablet. - Return the list if all tablets have <=30s lag. - Filter by replication lag: for each tablet, if the mean value without it is more than 0.7 of the mean value across all tablets, it is valid. - Make sure we return at least minNumTablets tablets (if there are enough one with only low replication lag). - If one tablet is removed, run above steps again in case there are two tablets with high replication lag. (It should cover most cases.) For example, lags of (5s, 10s, 15s, 120s) return the first three; lags of (30m, 35m, 40m, 45m) return all.

One thing to know about this code: vttablet also has a couple flags that impact the logic here:

  • unhealthy_threshold: if replication lag is higher than this, a tablet will be reported as unhealthy. The default for this is 2h, same as the discovery_high_replication_lag_minimum_serving here.
  • degraded_threshold: this is only used by vttablet for display. It should match discovery_low_replication_lag here, so the vttablet status display matches what vtgate will do of it.

func (*TabletHealth) DeepEqual 

func (th *TabletHealth) DeepEqual(other *TabletHealth) bool

DeepEqual compares two TabletHealth. Since we include protos, we need to use proto.Equal on these.

func (*TabletHealth) GetHostNameLevel 

func (th *TabletHealth) GetHostNameLevel(level int) string

GetHostNameLevel returns the specified hostname level. If the level does not exist it will pick the closest level. This seems unused but can be utilized by certain url formatting templates. See getTabletDebugURL for more details.

func (*TabletHealth) GetTabletHostPort 

func (th *TabletHealth) GetTabletHostPort() string

GetTabletHostPort formats a tablet host port address.

func (*TabletHealth) MarshalJSON added in v0.14.0 

func (th *TabletHealth) MarshalJSON() ([]byte, error)

type TabletPicker 

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

TabletPicker gives a simplified API for picking tablets.

func NewTabletPicker 

func NewTabletPicker(
	ctx context.Context,
	ts *topo.Server,
	cells []string,
	localCell, keyspace, shard, tabletTypesStr string,
	options TabletPickerOptions,
	ignoreTablets ...*topodatapb.TabletAlias,
) (*TabletPicker, error)

NewTabletPicker returns a TabletPicker.

func (*TabletPicker) GetMatchingTablets added in v0.11.0 

func (tp *TabletPicker) GetMatchingTablets(ctx context.Context) []*topo.TabletInfo

GetMatchingTablets returns a list of TabletInfo for healthy serving tablets that match the cells, keyspace, shard and tabletTypes for this TabletPicker.

func (*TabletPicker) PickForStreaming 

func (tp *TabletPicker) PickForStreaming(ctx context.Context) (*topodatapb.Tablet, error)

PickForStreaming picks a tablet that is healthy and serving. Selection is based on CellPreference. See prioritizeTablets for prioritization logic.

type TabletPickerCellPreference added in v0.17.0 

type TabletPickerCellPreference int
const (
	// PreferLocalWithAlias gives preference to the local cell first, then specified cells, if any.
	// This is the default when no other option is provided.
	TabletPickerCellPreference_PreferLocalWithAlias TabletPickerCellPreference = iota
	// OnlySpecified only picks tablets from the list of cells given.
	TabletPickerCellPreference_OnlySpecified
)

type TabletPickerOptions added in v0.17.0 

type TabletPickerOptions struct {
	CellPreference           string
	TabletOrder              string
	IncludeNonServingTablets bool
}

type TabletPickerTabletOrder added in v0.17.0 

type TabletPickerTabletOrder int
const (
	// All provided tablet types are given equal priority. This is the default.
	TabletPickerTabletOrder_Any TabletPickerTabletOrder = iota
	// Provided tablet types are expected to be prioritized in the given order.
	TabletPickerTabletOrder_InOrder
	InOrderHint = "in_order:"
)

type TabletStatsList 

type TabletStatsList []*TabletHealth

TabletStatsList is used for sorting.

func (TabletStatsList) Len 

func (tsl TabletStatsList) Len() int

Len is part of sort.Interface.

func (TabletStatsList) Less 

func (tsl TabletStatsList) Less(i, j int) bool

Less is part of sort.Interface

func (TabletStatsList) Swap 

func (tsl TabletStatsList) Swap(i, j int)

Swap is part of sort.Interface

type TabletsCacheStatus 

type TabletsCacheStatus struct {
	Cell         string
	Target       *querypb.Target
	TabletsStats TabletStatsList
}

TabletsCacheStatus is the current tablets for a cell/target.

func (*TabletsCacheStatus) StatusAsHTML 

func (tcs *TabletsCacheStatus) StatusAsHTML() safehtml.HTML

StatusAsHTML returns an HTML version of the status.

type TabletsCacheStatusList 

type TabletsCacheStatusList []*TabletsCacheStatus

TabletsCacheStatusList is used for sorting.

func (TabletsCacheStatusList) Len 

func (tcsl TabletsCacheStatusList) Len() int

Len is part of sort.Interface.

func (TabletsCacheStatusList) Less 

func (tcsl TabletsCacheStatusList) Less(i, j int) bool

Less is part of sort.Interface

func (TabletsCacheStatusList) Swap 

func (tcsl TabletsCacheStatusList) Swap(i, j int)

Swap is part of sort.Interface

type TopologyWatcher 

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

TopologyWatcher polls the topology periodically for changes to the set of tablets. When tablets are added / removed / modified, it calls the AddTablet / RemoveTablet interface appropriately.

func NewTopologyWatcher 

func NewTopologyWatcher(ctx context.Context, topoServer *topo.Server, hc HealthCheck, f TabletFilter, cell string, refreshInterval time.Duration, refreshKnownTablets bool, topoReadConcurrency int) *TopologyWatcher

NewTopologyWatcher returns a TopologyWatcher that monitors all the tablets in a cell, and reloads them as needed.

func (*TopologyWatcher) RefreshLag 

func (tw *TopologyWatcher) RefreshLag() time.Duration

RefreshLag returns the time since the last refresh.

func (*TopologyWatcher) Start 

func (tw *TopologyWatcher) Start()

Start starts the topology watcher.

func (*TopologyWatcher) Stop 

func (tw *TopologyWatcher) Stop()

Stop stops the watcher. It does not clean up the tablets added to HealthCheck.

func (*TopologyWatcher) TopoChecksum 

func (tw *TopologyWatcher) TopoChecksum() uint32

TopoChecksum returns the checksum of the current state of the topo.

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.