VictoriaMetrics/lib/protoparser/opentsdbhttp/streamparser.go
Aliaksandr Valialkin b275983403
lib/writeconcurrencylimiter: improve the logic behind -maxConcurrentInserts limit
Previously the -maxConcurrentInserts was limiting the number of established client connections,
which write data to VictoriaMetrics. Some of these connections could be idle.
Such connections do not consume big amounts of CPU and RAM, so there is a little sense in limiting
the number of such connections. So now the -maxConcurrentInserts command-line option
limits the number of concurrently executed insert requests, not including idle connections.

It is recommended removing -maxConcurrentInserts command-line option, since the default value
for this option should work good for most cases.
2023-01-06 22:07:16 -08:00

172 lines
4.6 KiB
Go

package opentsdbhttp
import (
"bufio"
"flag"
"fmt"
"io"
"net/http"
"sync"
"time"
"github.com/VictoriaMetrics/VictoriaMetrics/lib/bytesutil"
"github.com/VictoriaMetrics/VictoriaMetrics/lib/cgroup"
"github.com/VictoriaMetrics/VictoriaMetrics/lib/fasttime"
"github.com/VictoriaMetrics/VictoriaMetrics/lib/flagutil"
"github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/common"
"github.com/VictoriaMetrics/VictoriaMetrics/lib/writeconcurrencylimiter"
"github.com/VictoriaMetrics/metrics"
)
var (
maxInsertRequestSize = flagutil.NewBytes("opentsdbhttp.maxInsertRequestSize", 32*1024*1024, "The maximum size of OpenTSDB HTTP put request")
trimTimestamp = flag.Duration("opentsdbhttpTrimTimestamp", time.Millisecond, "Trim timestamps for OpenTSDB HTTP data to this duration. "+
"Minimum practical duration is 1ms. Higher duration (i.e. 1s) may be used for reducing disk space usage for timestamp data")
)
// ParseStream parses OpenTSDB http lines from req and calls callback for the parsed rows.
//
// The callback can be called concurrently multiple times for streamed data from req.
//
// callback shouldn't hold rows after returning.
func ParseStream(req *http.Request, callback func(rows []Row) error) error {
wcr := writeconcurrencylimiter.GetReader(req.Body)
defer writeconcurrencylimiter.PutReader(wcr)
r := io.Reader(req.Body)
readCalls.Inc()
if req.Header.Get("Content-Encoding") == "gzip" {
zr, err := common.GetGzipReader(r)
if err != nil {
readErrors.Inc()
return fmt.Errorf("cannot read gzipped http protocol data: %w", err)
}
defer common.PutGzipReader(zr)
r = zr
}
ctx := getStreamContext(r)
defer putStreamContext(ctx)
// Read the request in ctx.reqBuf
lr := io.LimitReader(ctx.br, int64(maxInsertRequestSize.N)+1)
reqLen, err := ctx.reqBuf.ReadFrom(lr)
if err != nil {
readErrors.Inc()
return fmt.Errorf("cannot read HTTP OpenTSDB request: %w", err)
}
if reqLen > int64(maxInsertRequestSize.N) {
readErrors.Inc()
return fmt.Errorf("too big HTTP OpenTSDB request; mustn't exceed `-opentsdbhttp.maxInsertRequestSize=%d` bytes", maxInsertRequestSize.N)
}
// Process the request synchronously, since there is no sense in processing a single request asynchronously.
// Sync code is easier to read and understand.
p := getJSONParser()
defer putJSONParser(p)
v, err := p.ParseBytes(ctx.reqBuf.B)
if err != nil {
unmarshalErrors.Inc()
return fmt.Errorf("cannot parse HTTP OpenTSDB json: %w", err)
}
rs := getRows()
defer putRows(rs)
rs.Unmarshal(v)
rows := rs.Rows
rowsRead.Add(len(rows))
// Fill in missing timestamps
currentTimestamp := int64(fasttime.UnixTimestamp())
for i := range rows {
r := &rows[i]
if r.Timestamp == 0 {
r.Timestamp = currentTimestamp
}
}
// Convert timestamps in seconds to milliseconds if needed.
// See http://opentsdb.net/docs/javadoc/net/opentsdb/core/Const.html#SECOND_MASK
for i := range rows {
r := &rows[i]
if r.Timestamp&secondMask == 0 {
r.Timestamp *= 1e3
}
}
// Trim timestamps if required.
if tsTrim := trimTimestamp.Milliseconds(); tsTrim > 1 {
for i := range rows {
row := &rows[i]
row.Timestamp -= row.Timestamp % tsTrim
}
}
if err := callback(rows); err != nil {
return fmt.Errorf("error when processing imported data: %w", err)
}
return nil
}
const secondMask int64 = 0x7FFFFFFF00000000
type streamContext struct {
br *bufio.Reader
reqBuf bytesutil.ByteBuffer
}
func (ctx *streamContext) reset() {
ctx.br.Reset(nil)
ctx.reqBuf.Reset()
}
var (
readCalls = metrics.NewCounter(`vm_protoparser_read_calls_total{type="opentsdbhttp"}`)
readErrors = metrics.NewCounter(`vm_protoparser_read_errors_total{type="opentsdbhttp"}`)
rowsRead = metrics.NewCounter(`vm_protoparser_rows_read_total{type="opentsdbhttp"}`)
unmarshalErrors = metrics.NewCounter(`vm_protoparser_unmarshal_errors_total{type="opentsdbhttp"}`)
)
func getStreamContext(r io.Reader) *streamContext {
select {
case ctx := <-streamContextPoolCh:
ctx.br.Reset(r)
return ctx
default:
if v := streamContextPool.Get(); v != nil {
ctx := v.(*streamContext)
ctx.br.Reset(r)
return ctx
}
return &streamContext{
br: bufio.NewReaderSize(r, 64*1024),
}
}
}
func putStreamContext(ctx *streamContext) {
ctx.reset()
select {
case streamContextPoolCh <- ctx:
default:
streamContextPool.Put(ctx)
}
}
var streamContextPool sync.Pool
var streamContextPoolCh = make(chan *streamContext, cgroup.AvailableCPUs())
func getRows() *Rows {
v := rowsPool.Get()
if v == nil {
return &Rows{}
}
return v.(*Rows)
}
func putRows(rs *Rows) {
rs.Reset()
rowsPool.Put(rs)
}
var rowsPool sync.Pool