func (self *CoordinatorImpl) shouldQuerySequentially(shards []*cluster.ShardData, querySpec *parser.QuerySpec) bool {
// if the query isn't a select, then it doesn't matter
if querySpec.SelectQuery != nil {
return false
}
// if the query is a regex, we can't predic the number of responses
// we get back
if querySpec.IsRegex() {
return true
}
groupByClause := querySpec.SelectQuery().GetGroupByClause()
// if there's no group by clause, then we're returning raw points
// with some math done on them, thus we can't predict the number of
// points
if groupByClause == nil {
return true
}
// if there's a group by clause but no group by interval, we can't
// predict the cardinality of the columns used in the group by
// interval, thus we can't predict the number of responses returned
// from the shard
if querySpec.GetGroupByInterval() == nil {
return true
}
// if there's a group by time and other columns, then the previous
// logic holds
if len(groupByClause.Elems) > 1 {
return true
}
// parallel querying only if we're querying a single series, with
// group by time only
return false
}
func (self *Shard) Query(querySpec *parser.QuerySpec, processor cluster.QueryProcessor) error {
if querySpec.IsListSeriesQuery() {
return self.executeListSeriesQuery(querySpec, processor)
} else if querySpec.IsDeleteFromSeriesQuery() {
return self.executeDeleteQuery(querySpec, processor)
}
seriesAndColumns := querySpec.SelectQuery().GetReferencedColumns()
if !self.hasReadAccess(querySpec) {
return errors.New("User does not have access to one or more of the series requested.")
}
for series, columns := range seriesAndColumns {
if regex, ok := series.GetCompiledRegex(); ok {
seriesNames := self.metaStore.GetSeriesForDatabaseAndRegex(querySpec.Database(), regex)
for _, name := range seriesNames {
if !querySpec.HasReadAccess(name) {
continue
}
err := self.executeQueryForSeries(querySpec, name, columns, processor)
if err != nil {
return err
}
}
} else {
err := self.executeQueryForSeries(querySpec, series.Name, columns, processor)
if err != nil {
return err
}
}
}
return nil
}
func (self *CoordinatorImpl) getShardsAndProcessor(querySpec *parser.QuerySpec, writer SeriesWriter) ([]*cluster.ShardData, cluster.QueryProcessor, chan bool, error) {
shards := self.clusterConfiguration.GetShards(querySpec)
shouldAggregateLocally := self.shouldAggregateLocally(shards, querySpec)
var err error
var processor cluster.QueryProcessor
responseChan := make(chan *protocol.Response)
seriesClosed := make(chan bool)
selectQuery := querySpec.SelectQuery()
if selectQuery != nil {
if !shouldAggregateLocally {
// if we should aggregate in the coordinator (i.e. aggregation
// isn't happening locally at the shard level), create an engine
processor, err = engine.NewQueryEngine(querySpec.SelectQuery(), responseChan)
} else {
// if we have a query with limit, then create an engine, or we can
// make the passthrough limit aware
processor = engine.NewPassthroughEngineWithLimit(responseChan, 100, selectQuery.Limit)
}
} else if !shouldAggregateLocally {
processor = engine.NewPassthroughEngine(responseChan, 100)
}
if err != nil {
return nil, nil, nil, err
}
if processor == nil {
return shards, nil, nil, nil
}
go func() {
for {
response := <-responseChan
if *response.Type == endStreamResponse || *response.Type == accessDeniedResponse {
writer.Close()
seriesClosed <- true
return
}
if !(*response.Type == queryResponse && querySpec.IsExplainQuery()) {
if response.Series != nil && len(response.Series.Points) > 0 {
writer.Write(response.Series)
}
}
}
}()
return shards, processor, seriesClosed, nil
}
func (self *CoordinatorImpl) checkPermission(user common.User, querySpec *parser.QuerySpec) error {
// if this isn't a regex query do the permission check here
fromClause := querySpec.SelectQuery().GetFromClause()
for _, n := range fromClause.Names {
if _, ok := n.Name.GetCompiledRegex(); ok {
break
} else if name := n.Name.Name; !user.HasReadAccess(name) {
return fmt.Errorf("User doesn't have read access to %s", name)
}
}
return nil
}
func (self *ShardData) Query(querySpec *parser.QuerySpec, response chan *protocol.Response) error {
// This is only for queries that are deletes or drops. They need to be sent everywhere as opposed to just the local or one of the remote shards.
// But this boolean should only be set to true on the server that receives the initial query.
if querySpec.RunAgainstAllServersInShard {
if querySpec.IsDeleteFromSeriesQuery() {
return self.logAndHandleDeleteQuery(querySpec, response)
} else if querySpec.IsDropSeriesQuery() {
return self.logAndHandleDropSeriesQuery(querySpec, response)
}
}
if self.localShard != nil {
var processor QueryProcessor
if querySpec.IsListSeriesQuery() {
processor = engine.NewListSeriesEngine(response)
} else if querySpec.IsDeleteFromSeriesQuery() || querySpec.IsDropSeriesQuery() || querySpec.IsSinglePointQuery() {
maxDeleteResults := 10000
processor = engine.NewPassthroughEngine(response, maxDeleteResults)
} else {
if self.ShouldAggregateLocally(querySpec) {
processor = engine.NewQueryEngine(querySpec.SelectQuery(), response)
} else {
maxPointsToBufferBeforeSending := 1000
processor = engine.NewPassthroughEngine(response, maxPointsToBufferBeforeSending)
}
}
err := self.localShard.Query(querySpec, processor)
processor.Close()
return err
}
healthyServers := make([]*ClusterServer, 0, len(self.clusterServers))
for _, s := range self.clusterServers {
if !s.IsUp() {
continue
}
healthyServers = append(healthyServers, s)
}
healthyCount := len(healthyServers)
if healthyCount == 0 {
message := fmt.Sprintf("No servers up to query shard %d", self.id)
response <- &protocol.Response{Type: &endStreamResponse, ErrorMessage: &message}
return errors.New(message)
}
randServerIndex := int(time.Now().UnixNano() % int64(healthyCount))
server := healthyServers[randServerIndex]
request := self.createRequest(querySpec)
return server.MakeRequest(request, response)
}
func (self *CoordinatorImpl) shouldQuerySequentially(shards []*cluster.ShardData, querySpec *parser.QuerySpec) bool {
// if the query isn't a select, then it doesn't matter
if querySpec.SelectQuery() == nil {
return false
}
// if the query is a regex, we can't predic the number of responses
// we get back
if querySpec.IsRegex() {
return true
}
groupByClause := querySpec.SelectQuery().GetGroupByClause()
// if there's no group by clause, then we're returning raw points
// with some math done on them, thus we can't predict the number of
// points
if groupByClause == nil {
return true
}
// if there's a group by clause but no group by interval, we can't
// predict the cardinality of the columns used in the group by
// interval, thus we can't predict the number of responses returned
// from the shard
if querySpec.GetGroupByInterval() == nil {
return true
}
// if there's a group by time and other columns, then the previous
// logic holds
if len(groupByClause.Elems) > 1 {
return true
}
if !self.shouldAggregateLocally(shards, querySpec) {
return true
}
for _, shard := range shards {
bufferSize := shard.QueryResponseBufferSize(querySpec, self.config.LevelDbPointBatchSize)
// if the number of repsonses is too big, do a sequential querying
if bufferSize > self.config.ClusterMaxResponseBufferSize {
return true
}
}
// parallel querying only if we're querying a single series, with
// group by time only
return false
}
func (self *Shard) fetchSinglePoint(querySpec *parser.QuerySpec, series string, fields []*metastore.Field) (*protocol.Series, error) {
query := querySpec.SelectQuery()
fieldCount := len(fields)
fieldNames := make([]string, 0, fieldCount)
point := &protocol.Point{Values: make([]*protocol.FieldValue, 0, fieldCount)}
timestamp := common.TimeToMicroseconds(query.GetStartTime())
sequenceNumber, err := query.GetSinglePointQuerySequenceNumber()
if err != nil {
return nil, err
}
timeAndSequenceBuffer := bytes.NewBuffer(make([]byte, 0, 16))
binary.Write(timeAndSequenceBuffer, binary.BigEndian, self.convertTimestampToUint(×tamp))
binary.Write(timeAndSequenceBuffer, binary.BigEndian, sequenceNumber)
sequenceNumber_uint64 := uint64(sequenceNumber)
point.SequenceNumber = &sequenceNumber_uint64
point.SetTimestampInMicroseconds(timestamp)
timeAndSequenceBytes := timeAndSequenceBuffer.Bytes()
for _, field := range fields {
pointKeyBuff := bytes.NewBuffer(make([]byte, 0, 24))
pointKeyBuff.Write(field.IdAsBytes())
pointKeyBuff.Write(timeAndSequenceBytes)
if data, err := self.db.Get(pointKeyBuff.Bytes()); err != nil {
return nil, err
} else {
fieldValue := &protocol.FieldValue{}
err := proto.Unmarshal(data, fieldValue)
if err != nil {
return nil, err
}
if data != nil {
fieldNames = append(fieldNames, field.Name)
point.Values = append(point.Values, fieldValue)
}
}
}
result := &protocol.Series{Name: &series, Fields: fieldNames, Points: []*protocol.Point{point}}
return result, nil
}
func (self *Shard) executeQueryForSeries(querySpec *parser.QuerySpec, seriesName string, columns []string, processor cluster.QueryProcessor) error {
startTimeBytes := self.byteArrayForTime(querySpec.GetStartTime())
endTimeBytes := self.byteArrayForTime(querySpec.GetEndTime())
fields, err := self.getFieldsForSeries(querySpec.Database(), seriesName, columns)
if err != nil {
// because a db is distributed across the cluster, it's possible we don't have the series indexed here. ignore
switch err := err.(type) {
case FieldLookupError:
log.Debug("Cannot find fields %v", columns)
return nil
default:
log.Error("Error looking up fields for %s: %s", seriesName, err)
return fmt.Errorf("Error looking up fields for %s: %s", seriesName, err)
}
}
fieldCount := len(fields)
rawColumnValues := make([]rawColumnValue, fieldCount, fieldCount)
query := querySpec.SelectQuery()
aliases := query.GetTableAliases(seriesName)
if querySpec.IsSinglePointQuery() {
series, err := self.fetchSinglePoint(querySpec, seriesName, fields)
if err != nil {
log.Error("Error reading a single point: %s", err)
return err
}
if len(series.Points) > 0 {
processor.YieldPoint(series.Name, series.Fields, series.Points[0])
}
return nil
}
fieldNames, iterators := self.getIterators(fields, startTimeBytes, endTimeBytes, query.Ascending)
defer func() {
for _, it := range iterators {
it.Close()
}
}()
seriesOutgoing := &protocol.Series{Name: protocol.String(seriesName), Fields: fieldNames, Points: make([]*protocol.Point, 0, self.pointBatchSize)}
// TODO: clean up, this is super gnarly
// optimize for the case where we're pulling back only a single column or aggregate
buffer := bytes.NewBuffer(nil)
valueBuffer := proto.NewBuffer(nil)
for {
isValid := false
point := &protocol.Point{Values: make([]*protocol.FieldValue, fieldCount, fieldCount)}
for i, it := range iterators {
if rawColumnValues[i].value != nil || !it.Valid() {
if err := it.Error(); err != nil {
return err
}
continue
}
key := it.Key()
if len(key) < 16 {
continue
}
if !isPointInRange(fields[i].Id, startTimeBytes, endTimeBytes, key) {
continue
}
value := it.Value()
sequenceNumber := key[16:]
rawTime := key[8:16]
rawColumnValues[i] = rawColumnValue{time: rawTime, sequence: sequenceNumber, value: value}
}
var pointTimeRaw []byte
var pointSequenceRaw []byte
// choose the highest (or lowest in case of ascending queries) timestamp
// and sequence number. that will become the timestamp and sequence of
// the next point.
for _, value := range rawColumnValues {
if value.value == nil {
continue
}
pointTimeRaw, pointSequenceRaw = value.updatePointTimeAndSequence(pointTimeRaw,
pointSequenceRaw, query.Ascending)
}
for i, iterator := range iterators {
// if the value is nil or doesn't match the point's timestamp and sequence number
// then skip it
if rawColumnValues[i].value == nil ||
!bytes.Equal(rawColumnValues[i].time, pointTimeRaw) ||
!bytes.Equal(rawColumnValues[i].sequence, pointSequenceRaw) {
point.Values[i] = &protocol.FieldValue{IsNull: &TRUE}
continue
}
// if we emitted at lease one column, then we should keep
// trying to get more points
isValid = true
// advance the iterator to read a new value in the next iteration
if query.Ascending {
iterator.Next()
} else {
iterator.Prev()
}
fv := &protocol.FieldValue{}
valueBuffer.SetBuf(rawColumnValues[i].value)
err := valueBuffer.Unmarshal(fv)
if err != nil {
log.Error("Error while running query: %s", err)
return err
}
point.Values[i] = fv
rawColumnValues[i].value = nil
}
var sequence uint64
var t uint64
// set the point sequence number and timestamp
buffer.Reset()
buffer.Write(pointSequenceRaw)
binary.Read(buffer, binary.BigEndian, &sequence)
buffer.Reset()
buffer.Write(pointTimeRaw)
binary.Read(buffer, binary.BigEndian, &t)
time := self.convertUintTimestampToInt64(&t)
point.SetTimestampInMicroseconds(time)
point.SequenceNumber = &sequence
// stop the loop if we ran out of points
if !isValid {
break
}
shouldContinue := true
seriesOutgoing.Points = append(seriesOutgoing.Points, point)
if len(seriesOutgoing.Points) >= self.pointBatchSize {
for _, alias := range aliases {
series := &protocol.Series{
Name: proto.String(alias),
Fields: fieldNames,
Points: seriesOutgoing.Points,
}
if !processor.YieldSeries(series) {
log.Info("Stopping processing")
shouldContinue = false
}
}
seriesOutgoing = &protocol.Series{Name: protocol.String(seriesName), Fields: fieldNames, Points: make([]*protocol.Point, 0, self.pointBatchSize)}
}
if !shouldContinue {
break
}
}
//Yield remaining data
for _, alias := range aliases {
log.Debug("Final Flush %s", alias)
series := &protocol.Series{Name: protocol.String(alias), Fields: seriesOutgoing.Fields, Points: seriesOutgoing.Points}
if !processor.YieldSeries(series) {
log.Debug("Cancelled...")
}
}
log.Debug("Finished running query %s", query.GetQueryString())
return nil
}
func (self *ShardData) Query(querySpec *parser.QuerySpec, response chan *p.Response) {
log.Debug("QUERY: shard %d, query '%s'", self.Id(), querySpec.GetQueryString())
defer common.RecoverFunc(querySpec.Database(), querySpec.GetQueryString(), func(err interface{}) {
response <- &p.Response{Type: &endStreamResponse, ErrorMessage: p.String(fmt.Sprintf("%s", err))}
})
// This is only for queries that are deletes or drops. They need to be sent everywhere as opposed to just the local or one of the remote shards.
// But this boolean should only be set to true on the server that receives the initial query.
if querySpec.RunAgainstAllServersInShard {
if querySpec.IsDeleteFromSeriesQuery() {
self.logAndHandleDeleteQuery(querySpec, response)
} else if querySpec.IsDropSeriesQuery() {
self.logAndHandleDropSeriesQuery(querySpec, response)
}
}
if self.IsLocal {
var processor QueryProcessor
var err error
if querySpec.IsListSeriesQuery() {
processor = engine.NewListSeriesEngine(response)
} else if querySpec.IsDeleteFromSeriesQuery() || querySpec.IsDropSeriesQuery() || querySpec.IsSinglePointQuery() {
maxDeleteResults := 10000
processor = engine.NewPassthroughEngine(response, maxDeleteResults)
} else {
query := querySpec.SelectQuery()
if self.ShouldAggregateLocally(querySpec) {
log.Debug("creating a query engine")
processor, err = engine.NewQueryEngine(query, response)
if err != nil {
response <- &p.Response{Type: &endStreamResponse, ErrorMessage: p.String(err.Error())}
log.Error("Error while creating engine: %s", err)
return
}
processor.SetShardInfo(int(self.Id()), self.IsLocal)
} else if query.HasAggregates() {
maxPointsToBufferBeforeSending := 1000
log.Debug("creating a passthrough engine")
processor = engine.NewPassthroughEngine(response, maxPointsToBufferBeforeSending)
} else {
maxPointsToBufferBeforeSending := 1000
log.Debug("creating a passthrough engine with limit")
processor = engine.NewPassthroughEngineWithLimit(response, maxPointsToBufferBeforeSending, query.Limit)
}
if query.GetFromClause().Type != parser.FromClauseInnerJoin {
// Joins do their own filtering since we need to get all
// points before filtering. This is due to the fact that some
// where expressions will be difficult to compute before the
// points are joined together, think where clause with
// left.column = 'something' or right.column =
// 'something_else'. We can't filter the individual series
// separately. The filtering happens in merge.go:55
processor = engine.NewFilteringEngine(query, processor)
}
}
shard, err := self.store.GetOrCreateShard(self.id)
if err != nil {
response <- &p.Response{Type: &endStreamResponse, ErrorMessage: p.String(err.Error())}
log.Error("Error while getting shards: %s", err)
return
}
defer self.store.ReturnShard(self.id)
err = shard.Query(querySpec, processor)
// if we call Close() in case of an error it will mask the error
if err != nil {
response <- &p.Response{Type: &endStreamResponse, ErrorMessage: p.String(err.Error())}
return
}
processor.Close()
response <- &p.Response{Type: &endStreamResponse}
return
}
if server := self.randomHealthyServer(); server != nil {
log.Debug("Querying server %d for shard %d", server.GetId(), self.Id())
request := self.createRequest(querySpec)
server.MakeRequest(request, response)
return
}
message := fmt.Sprintf("No servers up to query shard %d", self.id)
response <- &p.Response{Type: &endStreamResponse, ErrorMessage: &message}
log.Error(message)
}
func (self *ShardData) Query(querySpec *parser.QuerySpec, response chan *p.Response) {
// This is only for queries that are deletes or drops. They need to be sent everywhere as opposed to just the local or one of the remote shards.
// But this boolean should only be set to true on the server that receives the initial query.
if querySpec.RunAgainstAllServersInShard {
if querySpec.IsDeleteFromSeriesQuery() {
self.logAndHandleDeleteQuery(querySpec, response)
} else if querySpec.IsDropSeriesQuery() {
self.logAndHandleDropSeriesQuery(querySpec, response)
}
}
if self.IsLocal {
var processor QueryProcessor
var err error
if querySpec.IsListSeriesQuery() {
processor = engine.NewListSeriesEngine(response)
} else if querySpec.IsDeleteFromSeriesQuery() || querySpec.IsDropSeriesQuery() || querySpec.IsSinglePointQuery() {
maxDeleteResults := 10000
processor = engine.NewPassthroughEngine(response, maxDeleteResults)
} else {
query := querySpec.SelectQuery()
if self.ShouldAggregateLocally(querySpec) {
log.Debug("creating a query engine\n")
processor, err = engine.NewQueryEngine(query, response)
if err != nil {
response <- &p.Response{Type: &endStreamResponse, ErrorMessage: p.String(err.Error())}
log.Error("Error while creating engine: %s", err)
return
}
processor.SetShardInfo(int(self.Id()), self.IsLocal)
} else if query.HasAggregates() {
maxPointsToBufferBeforeSending := 1000
log.Debug("creating a passthrough engine\n")
processor = engine.NewPassthroughEngine(response, maxPointsToBufferBeforeSending)
} else {
maxPointsToBufferBeforeSending := 1000
log.Debug("creating a passthrough engine with limit\n")
processor = engine.NewPassthroughEngineWithLimit(response, maxPointsToBufferBeforeSending, query.Limit)
}
processor = engine.NewFilteringEngine(query, processor)
}
shard, err := self.store.GetOrCreateShard(self.id)
if err != nil {
response <- &p.Response{Type: &endStreamResponse, ErrorMessage: p.String(err.Error())}
log.Error("Error while getting shards: %s", err)
return
}
defer self.store.ReturnShard(self.id)
err = shard.Query(querySpec, processor)
processor.Close()
if err != nil {
response <- &p.Response{Type: &endStreamResponse, ErrorMessage: p.String(err.Error())}
}
response <- &p.Response{Type: &endStreamResponse}
return
}
healthyServers := make([]*ClusterServer, 0, len(self.clusterServers))
for _, s := range self.clusterServers {
if !s.IsUp() {
continue
}
healthyServers = append(healthyServers, s)
}
healthyCount := len(healthyServers)
if healthyCount == 0 {
message := fmt.Sprintf("No servers up to query shard %d", self.id)
response <- &p.Response{Type: &endStreamResponse, ErrorMessage: &message}
log.Error(message)
return
}
randServerIndex := int(time.Now().UnixNano() % int64(healthyCount))
server := healthyServers[randServerIndex]
request := self.createRequest(querySpec)
server.MakeRequest(request, response)
}
func (self *CoordinatorImpl) runQuerySpec(querySpec *parser.QuerySpec, seriesWriter SeriesWriter) error {
shards := self.clusterConfiguration.GetShards(querySpec)
shouldAggregateLocally := true
var processor cluster.QueryProcessor
var responseChan chan *protocol.Response
var seriesClosed chan bool
for _, s := range shards {
// If the aggregation is done at the shard level, we don't need to
// do it here at the coordinator level.
if !s.ShouldAggregateLocally(querySpec) {
seriesClosed = make(chan bool)
shouldAggregateLocally = false
responseChan = make(chan *protocol.Response)
if querySpec.SelectQuery() != nil {
processor = engine.NewQueryEngine(querySpec.SelectQuery(), responseChan)
} else {
bufferSize := 100
processor = engine.NewPassthroughEngine(responseChan, bufferSize)
}
go func() {
for {
res := <-responseChan
if *res.Type == endStreamResponse || *res.Type == accessDeniedResponse {
seriesWriter.Close()
seriesClosed <- true
return
}
if res.Series != nil && len(res.Series.Points) > 0 {
seriesWriter.Write(res.Series)
}
}
}()
break
}
}
responses := make([]chan *protocol.Response, 0)
for _, shard := range shards {
responseChan := make(chan *protocol.Response, self.config.QueryShardBufferSize)
go shard.Query(querySpec, responseChan)
responses = append(responses, responseChan)
}
for i, responseChan := range responses {
log.Debug("READING: shard: ", shards[i].String())
for {
response := <-responseChan
log.Debug("GOT RESPONSE: ", response.Type, response.Series)
if *response.Type == endStreamResponse || *response.Type == accessDeniedResponse {
break
}
if shouldAggregateLocally {
log.Debug("WRITING: ", len(response.Series.Points))
seriesWriter.Write(response.Series)
log.Debug("WRITING (done)")
continue
}
// if the data wasn't aggregated at the shard level, aggregate
// the data here
log.Debug("YIELDING: ", len(response.Series.Points))
if response.Series != nil {
for _, p := range response.Series.Points {
processor.YieldPoint(response.Series.Name, response.Series.Fields, p)
}
}
}
log.Debug("DONE: shard: ", shards[i].String())
}
if !shouldAggregateLocally {
processor.Close()
<-seriesClosed
return nil
}
seriesWriter.Close()
return nil
}