prometheus/tsdb/chunks/head_chunks_test.go

279 lines
8.3 KiB
Go
Raw Normal View History

// Copyright 2020 The Prometheus Authors
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
package chunks
import (
"encoding/binary"
"io/ioutil"
"math/rand"
"os"
"testing"
"time"
"github.com/prometheus/prometheus/tsdb/chunkenc"
"github.com/prometheus/prometheus/util/testutil"
)
func TestHeadReadWriter_WriteChunk_Chunk_IterateChunks(t *testing.T) {
hrw, close := testHeadReadWriter(t)
defer func() {
testutil.Ok(t, hrw.Close())
close()
}()
expectedBytes := []byte{}
nextChunkOffset := uint64(HeadChunkFileHeaderSize)
chkCRC32 := newCRC32()
type expectedDataType struct {
seriesRef, chunkRef uint64
mint, maxt int64
M-map full chunks of Head from disk (#6679) When appending to the head and a chunk is full it is flushed to the disk and m-mapped (memory mapped) to free up memory Prom startup now happens in these stages - Iterate the m-maped chunks from disk and keep a map of series reference to its slice of mmapped chunks. - Iterate the WAL as usual. Whenever we create a new series, look for it's mmapped chunks in the map created before and add it to that series. If a head chunk is corrupted the currpted one and all chunks after that are deleted and the data after the corruption is recovered from the existing WAL which means that a corruption in m-mapped files results in NO data loss. [Mmaped chunks format](https://github.com/prometheus/prometheus/blob/master/tsdb/docs/format/head_chunks.md) - main difference is that the chunk for mmaping now also includes series reference because there is no index for mapping series to chunks. [The block chunks](https://github.com/prometheus/prometheus/blob/master/tsdb/docs/format/chunks.md) are accessed from the index which includes the offsets for the chunks in the chunks file - example - chunks of series ID have offsets 200, 500 etc in the chunk files. In case of mmaped chunks, the offsets are stored in memory and accessed from that. During WAL replay, these offsets are restored by iterating all m-mapped chunks as stated above by matching the series id present in the chunk header and offset of that chunk in that file. **Prombench results** _WAL Replay_ 1h Wal reply time 30% less wal reply time - 4m31 vs 3m36 2h Wal reply time 20% less wal reply time - 8m16 vs 7m _Memory During WAL Replay_ High Churn: 10-15% less RAM - 32gb vs 28gb 20% less RAM after compaction 34gb vs 27gb No Churn: 20-30% less RAM - 23gb vs 18gb 40% less RAM after compaction 32.5gb vs 20gb Screenshots are in [this comment](https://github.com/prometheus/prometheus/pull/6679#issuecomment-621678932) Signed-off-by: Ganesh Vernekar <cs15btech11018@iith.ac.in>
2020-05-06 08:30:00 -07:00
numSamples uint16
chunk chunkenc.Chunk
}
expectedData := []expectedDataType{}
var buf [MaxHeadChunkMetaSize]byte
totalChunks := 0
var firstFileName string
for hrw.curFileSequence < 3 || hrw.chkWriter.Buffered() == 0 {
for i := 0; i < 100; i++ {
seriesRef, chkRef, mint, maxt, chunk := createChunk(t, totalChunks, hrw)
totalChunks++
expectedData = append(expectedData, expectedDataType{
M-map full chunks of Head from disk (#6679) When appending to the head and a chunk is full it is flushed to the disk and m-mapped (memory mapped) to free up memory Prom startup now happens in these stages - Iterate the m-maped chunks from disk and keep a map of series reference to its slice of mmapped chunks. - Iterate the WAL as usual. Whenever we create a new series, look for it's mmapped chunks in the map created before and add it to that series. If a head chunk is corrupted the currpted one and all chunks after that are deleted and the data after the corruption is recovered from the existing WAL which means that a corruption in m-mapped files results in NO data loss. [Mmaped chunks format](https://github.com/prometheus/prometheus/blob/master/tsdb/docs/format/head_chunks.md) - main difference is that the chunk for mmaping now also includes series reference because there is no index for mapping series to chunks. [The block chunks](https://github.com/prometheus/prometheus/blob/master/tsdb/docs/format/chunks.md) are accessed from the index which includes the offsets for the chunks in the chunks file - example - chunks of series ID have offsets 200, 500 etc in the chunk files. In case of mmaped chunks, the offsets are stored in memory and accessed from that. During WAL replay, these offsets are restored by iterating all m-mapped chunks as stated above by matching the series id present in the chunk header and offset of that chunk in that file. **Prombench results** _WAL Replay_ 1h Wal reply time 30% less wal reply time - 4m31 vs 3m36 2h Wal reply time 20% less wal reply time - 8m16 vs 7m _Memory During WAL Replay_ High Churn: 10-15% less RAM - 32gb vs 28gb 20% less RAM after compaction 34gb vs 27gb No Churn: 20-30% less RAM - 23gb vs 18gb 40% less RAM after compaction 32.5gb vs 20gb Screenshots are in [this comment](https://github.com/prometheus/prometheus/pull/6679#issuecomment-621678932) Signed-off-by: Ganesh Vernekar <cs15btech11018@iith.ac.in>
2020-05-06 08:30:00 -07:00
seriesRef: seriesRef,
mint: mint,
maxt: maxt,
chunkRef: chkRef,
chunk: chunk,
numSamples: uint16(chunk.NumSamples()),
})
if hrw.curFileSequence != 1 {
// We are checking for bytes written only for the first file.
continue
}
// Calculating expected bytes written on disk for first file.
firstFileName = hrw.curFile.Name()
testutil.Equals(t, chunkRef(1, nextChunkOffset), chkRef)
bytesWritten := 0
chkCRC32.Reset()
binary.BigEndian.PutUint64(buf[bytesWritten:], seriesRef)
bytesWritten += SeriesRefSize
binary.BigEndian.PutUint64(buf[bytesWritten:], uint64(mint))
bytesWritten += MintMaxtSize
binary.BigEndian.PutUint64(buf[bytesWritten:], uint64(maxt))
bytesWritten += MintMaxtSize
buf[bytesWritten] = byte(chunk.Encoding())
bytesWritten += ChunkEncodingSize
n := binary.PutUvarint(buf[bytesWritten:], uint64(len(chunk.Bytes())))
bytesWritten += n
expectedBytes = append(expectedBytes, buf[:bytesWritten]...)
_, err := chkCRC32.Write(buf[:bytesWritten])
testutil.Ok(t, err)
expectedBytes = append(expectedBytes, chunk.Bytes()...)
_, err = chkCRC32.Write(chunk.Bytes())
testutil.Ok(t, err)
expectedBytes = append(expectedBytes, chkCRC32.Sum(nil)...)
// += seriesRef, mint, maxt, encoding, chunk data len, chunk data, CRC.
nextChunkOffset += SeriesRefSize + 2*MintMaxtSize + ChunkEncodingSize + uint64(n) + uint64(len(chunk.Bytes())) + CRCSize
}
}
// Checking on-disk bytes for the first file.
testutil.Assert(t, len(hrw.mmappedChunkFiles) == 3 && len(hrw.closers) == 3, "expected 3 mmapped files, got %d", len(hrw.mmappedChunkFiles))
actualBytes, err := ioutil.ReadFile(firstFileName)
testutil.Ok(t, err)
// Check header of the segment file.
testutil.Equals(t, MagicHeadChunks, int(binary.BigEndian.Uint32(actualBytes[0:MagicChunksSize])))
testutil.Equals(t, chunksFormatV1, int(actualBytes[MagicChunksSize]))
// Remaining chunk data.
fileEnd := HeadChunkFileHeaderSize + len(expectedBytes)
testutil.Equals(t, expectedBytes, actualBytes[HeadChunkFileHeaderSize:fileEnd])
// Test for the next chunk header to be all 0s. That marks the end of the file.
for _, b := range actualBytes[fileEnd : fileEnd+MaxHeadChunkMetaSize] {
testutil.Equals(t, byte(0), b)
}
// Testing reading of chunks.
for _, exp := range expectedData {
actChunk, err := hrw.Chunk(exp.chunkRef)
testutil.Ok(t, err)
testutil.Equals(t, exp.chunk.Bytes(), actChunk.Bytes())
}
// Testing IterateAllChunks method.
dir := hrw.dir.Name()
testutil.Ok(t, hrw.Close())
hrw, err = NewChunkDiskMapper(dir, chunkenc.NewPool())
testutil.Ok(t, err)
idx := 0
M-map full chunks of Head from disk (#6679) When appending to the head and a chunk is full it is flushed to the disk and m-mapped (memory mapped) to free up memory Prom startup now happens in these stages - Iterate the m-maped chunks from disk and keep a map of series reference to its slice of mmapped chunks. - Iterate the WAL as usual. Whenever we create a new series, look for it's mmapped chunks in the map created before and add it to that series. If a head chunk is corrupted the currpted one and all chunks after that are deleted and the data after the corruption is recovered from the existing WAL which means that a corruption in m-mapped files results in NO data loss. [Mmaped chunks format](https://github.com/prometheus/prometheus/blob/master/tsdb/docs/format/head_chunks.md) - main difference is that the chunk for mmaping now also includes series reference because there is no index for mapping series to chunks. [The block chunks](https://github.com/prometheus/prometheus/blob/master/tsdb/docs/format/chunks.md) are accessed from the index which includes the offsets for the chunks in the chunks file - example - chunks of series ID have offsets 200, 500 etc in the chunk files. In case of mmaped chunks, the offsets are stored in memory and accessed from that. During WAL replay, these offsets are restored by iterating all m-mapped chunks as stated above by matching the series id present in the chunk header and offset of that chunk in that file. **Prombench results** _WAL Replay_ 1h Wal reply time 30% less wal reply time - 4m31 vs 3m36 2h Wal reply time 20% less wal reply time - 8m16 vs 7m _Memory During WAL Replay_ High Churn: 10-15% less RAM - 32gb vs 28gb 20% less RAM after compaction 34gb vs 27gb No Churn: 20-30% less RAM - 23gb vs 18gb 40% less RAM after compaction 32.5gb vs 20gb Screenshots are in [this comment](https://github.com/prometheus/prometheus/pull/6679#issuecomment-621678932) Signed-off-by: Ganesh Vernekar <cs15btech11018@iith.ac.in>
2020-05-06 08:30:00 -07:00
err = hrw.IterateAllChunks(func(seriesRef, chunkRef uint64, mint, maxt int64, numSamples uint16) error {
t.Helper()
expData := expectedData[idx]
testutil.Equals(t, expData.seriesRef, seriesRef)
testutil.Equals(t, expData.chunkRef, chunkRef)
testutil.Equals(t, expData.maxt, maxt)
testutil.Equals(t, expData.maxt, maxt)
M-map full chunks of Head from disk (#6679) When appending to the head and a chunk is full it is flushed to the disk and m-mapped (memory mapped) to free up memory Prom startup now happens in these stages - Iterate the m-maped chunks from disk and keep a map of series reference to its slice of mmapped chunks. - Iterate the WAL as usual. Whenever we create a new series, look for it's mmapped chunks in the map created before and add it to that series. If a head chunk is corrupted the currpted one and all chunks after that are deleted and the data after the corruption is recovered from the existing WAL which means that a corruption in m-mapped files results in NO data loss. [Mmaped chunks format](https://github.com/prometheus/prometheus/blob/master/tsdb/docs/format/head_chunks.md) - main difference is that the chunk for mmaping now also includes series reference because there is no index for mapping series to chunks. [The block chunks](https://github.com/prometheus/prometheus/blob/master/tsdb/docs/format/chunks.md) are accessed from the index which includes the offsets for the chunks in the chunks file - example - chunks of series ID have offsets 200, 500 etc in the chunk files. In case of mmaped chunks, the offsets are stored in memory and accessed from that. During WAL replay, these offsets are restored by iterating all m-mapped chunks as stated above by matching the series id present in the chunk header and offset of that chunk in that file. **Prombench results** _WAL Replay_ 1h Wal reply time 30% less wal reply time - 4m31 vs 3m36 2h Wal reply time 20% less wal reply time - 8m16 vs 7m _Memory During WAL Replay_ High Churn: 10-15% less RAM - 32gb vs 28gb 20% less RAM after compaction 34gb vs 27gb No Churn: 20-30% less RAM - 23gb vs 18gb 40% less RAM after compaction 32.5gb vs 20gb Screenshots are in [this comment](https://github.com/prometheus/prometheus/pull/6679#issuecomment-621678932) Signed-off-by: Ganesh Vernekar <cs15btech11018@iith.ac.in>
2020-05-06 08:30:00 -07:00
testutil.Equals(t, expData.numSamples, numSamples)
actChunk, err := hrw.Chunk(expData.chunkRef)
testutil.Ok(t, err)
testutil.Equals(t, expData.chunk.Bytes(), actChunk.Bytes())
idx++
return nil
})
testutil.Ok(t, err)
testutil.Equals(t, len(expectedData), idx)
}
func TestHeadReadWriter_Truncate(t *testing.T) {
hrw, close := testHeadReadWriter(t)
defer func() {
testutil.Ok(t, hrw.Close())
close()
}()
testutil.Assert(t, !hrw.fileMaxtSet, "")
M-map full chunks of Head from disk (#6679) When appending to the head and a chunk is full it is flushed to the disk and m-mapped (memory mapped) to free up memory Prom startup now happens in these stages - Iterate the m-maped chunks from disk and keep a map of series reference to its slice of mmapped chunks. - Iterate the WAL as usual. Whenever we create a new series, look for it's mmapped chunks in the map created before and add it to that series. If a head chunk is corrupted the currpted one and all chunks after that are deleted and the data after the corruption is recovered from the existing WAL which means that a corruption in m-mapped files results in NO data loss. [Mmaped chunks format](https://github.com/prometheus/prometheus/blob/master/tsdb/docs/format/head_chunks.md) - main difference is that the chunk for mmaping now also includes series reference because there is no index for mapping series to chunks. [The block chunks](https://github.com/prometheus/prometheus/blob/master/tsdb/docs/format/chunks.md) are accessed from the index which includes the offsets for the chunks in the chunks file - example - chunks of series ID have offsets 200, 500 etc in the chunk files. In case of mmaped chunks, the offsets are stored in memory and accessed from that. During WAL replay, these offsets are restored by iterating all m-mapped chunks as stated above by matching the series id present in the chunk header and offset of that chunk in that file. **Prombench results** _WAL Replay_ 1h Wal reply time 30% less wal reply time - 4m31 vs 3m36 2h Wal reply time 20% less wal reply time - 8m16 vs 7m _Memory During WAL Replay_ High Churn: 10-15% less RAM - 32gb vs 28gb 20% less RAM after compaction 34gb vs 27gb No Churn: 20-30% less RAM - 23gb vs 18gb 40% less RAM after compaction 32.5gb vs 20gb Screenshots are in [this comment](https://github.com/prometheus/prometheus/pull/6679#issuecomment-621678932) Signed-off-by: Ganesh Vernekar <cs15btech11018@iith.ac.in>
2020-05-06 08:30:00 -07:00
testutil.Ok(t, hrw.IterateAllChunks(func(_, _ uint64, _, _ int64, _ uint16) error { return nil }))
testutil.Assert(t, hrw.fileMaxtSet, "")
timeRange := 0
fileTimeStep := 100
totalFiles, after1stTruncation, after2ndTruncation := 7, 5, 3
var timeToTruncate, timeToTruncateAfterRestart int64
cutFile := func(i int) {
testutil.Ok(t, hrw.cut(int64(timeRange)))
mint := timeRange + 1 // Just after the the new file cut.
maxt := timeRange + fileTimeStep - 1 // Just before the next file.
// Write a chunks to set maxt for the segment.
_, err := hrw.WriteChunk(1, int64(mint), int64(maxt), randomChunk(t))
testutil.Ok(t, err)
if i == totalFiles-after1stTruncation+1 {
// Truncate the segment files before the 5th segment.
timeToTruncate = int64(mint)
} else if i == totalFiles-after2ndTruncation+1 {
// Truncate the segment files before the 3rd segment after restart.
timeToTruncateAfterRestart = int64(mint)
}
timeRange += fileTimeStep
}
// Cut segments.
for i := 1; i <= totalFiles; i++ {
cutFile(i)
}
// Verifying the the remaining files.
verifyRemainingFiles := func(remainingFiles int) {
t.Helper()
files, err := ioutil.ReadDir(hrw.dir.Name())
testutil.Ok(t, err)
testutil.Equals(t, remainingFiles, len(files))
testutil.Equals(t, remainingFiles, len(hrw.mmappedChunkFiles))
testutil.Equals(t, remainingFiles, len(hrw.closers))
for i := 1; i <= totalFiles; i++ {
_, ok := hrw.mmappedChunkFiles[i]
if i < totalFiles-remainingFiles+1 {
testutil.Equals(t, false, ok)
} else {
testutil.Equals(t, true, ok)
}
}
}
// Verify the number of segments.
verifyRemainingFiles(totalFiles)
// Truncating files.
testutil.Ok(t, hrw.Truncate(timeToTruncate))
verifyRemainingFiles(after1stTruncation)
dir := hrw.dir.Name()
testutil.Ok(t, hrw.Close())
// Restarted.
var err error
hrw, err = NewChunkDiskMapper(dir, chunkenc.NewPool())
testutil.Ok(t, err)
testutil.Assert(t, !hrw.fileMaxtSet, "")
M-map full chunks of Head from disk (#6679) When appending to the head and a chunk is full it is flushed to the disk and m-mapped (memory mapped) to free up memory Prom startup now happens in these stages - Iterate the m-maped chunks from disk and keep a map of series reference to its slice of mmapped chunks. - Iterate the WAL as usual. Whenever we create a new series, look for it's mmapped chunks in the map created before and add it to that series. If a head chunk is corrupted the currpted one and all chunks after that are deleted and the data after the corruption is recovered from the existing WAL which means that a corruption in m-mapped files results in NO data loss. [Mmaped chunks format](https://github.com/prometheus/prometheus/blob/master/tsdb/docs/format/head_chunks.md) - main difference is that the chunk for mmaping now also includes series reference because there is no index for mapping series to chunks. [The block chunks](https://github.com/prometheus/prometheus/blob/master/tsdb/docs/format/chunks.md) are accessed from the index which includes the offsets for the chunks in the chunks file - example - chunks of series ID have offsets 200, 500 etc in the chunk files. In case of mmaped chunks, the offsets are stored in memory and accessed from that. During WAL replay, these offsets are restored by iterating all m-mapped chunks as stated above by matching the series id present in the chunk header and offset of that chunk in that file. **Prombench results** _WAL Replay_ 1h Wal reply time 30% less wal reply time - 4m31 vs 3m36 2h Wal reply time 20% less wal reply time - 8m16 vs 7m _Memory During WAL Replay_ High Churn: 10-15% less RAM - 32gb vs 28gb 20% less RAM after compaction 34gb vs 27gb No Churn: 20-30% less RAM - 23gb vs 18gb 40% less RAM after compaction 32.5gb vs 20gb Screenshots are in [this comment](https://github.com/prometheus/prometheus/pull/6679#issuecomment-621678932) Signed-off-by: Ganesh Vernekar <cs15btech11018@iith.ac.in>
2020-05-06 08:30:00 -07:00
testutil.Ok(t, hrw.IterateAllChunks(func(_, _ uint64, _, _ int64, _ uint16) error { return nil }))
testutil.Assert(t, hrw.fileMaxtSet, "")
// Truncating files after restart.
testutil.Ok(t, hrw.Truncate(timeToTruncateAfterRestart))
verifyRemainingFiles(after2ndTruncation)
// Add another file to have an active file.
totalFiles++
cutFile(totalFiles)
// Truncating till current time should not delete the current active file.
testutil.Ok(t, hrw.Truncate(time.Now().UnixNano()/1e6))
verifyRemainingFiles(1)
}
func testHeadReadWriter(t *testing.T) (hrw *ChunkDiskMapper, close func()) {
tmpdir, err := ioutil.TempDir("", "data")
testutil.Ok(t, err)
hrw, err = NewChunkDiskMapper(tmpdir, chunkenc.NewPool())
testutil.Ok(t, err)
return hrw, func() {
testutil.Ok(t, os.RemoveAll(tmpdir))
}
}
func randomChunk(t *testing.T) chunkenc.Chunk {
chunk := chunkenc.NewXORChunk()
len := rand.Int() % 120
app, err := chunk.Appender()
testutil.Ok(t, err)
for i := 0; i < len; i++ {
app.Append(rand.Int63(), rand.Float64())
}
return chunk
}
func createChunk(t *testing.T, idx int, hrw *ChunkDiskMapper) (seriesRef uint64, chunkRef uint64, mint, maxt int64, chunk chunkenc.Chunk) {
var err error
seriesRef = uint64(rand.Int63())
mint = int64((idx)*1000 + 1)
maxt = int64((idx + 1) * 1000)
chunk = randomChunk(t)
chunkRef, err = hrw.WriteChunk(seriesRef, mint, maxt, chunk)
testutil.Ok(t, err)
return
}