VictoriaMetrics/lib/mergeset/table_test.go
Zakhar Bessarab 8198e7241d
lib/mergeset: add sparse indexdb cache (#7269)
Related issue:
https://github.com/VictoriaMetrics/VictoriaMetrics/issues/7182

- add a separate index cache for searches which might read through large
amounts of random entries. Primary use-case for this is retention and
downsampling filters, when applying filters background merge needs to
fetch large amount of random entries which pollutes an index cache.
Using different caches allows to reduce effect on memory usage and cache
efficiency of the main cache while still having high cache hit rate. A
separate cache size is 5% of allowed memory.

- reduce size of indexdb/dataBlocks cache in order to free memory for
new sparse cache. Reduced size by 5% and moved this to a separate cache.

- add a separate metricName search which does not cache metric names -
this is needed in order to allow disabling metric name caching when
applying downsampling/retention filters. Applying filters during
background merge accesses random entries, this fills up cache and does
not provide an actual improvement due to random access nature.

Merge performance and memory usage stats before and after the change:

- before

![image](https://github.com/user-attachments/assets/485fffbb-c225-47ae-b5c5-bc8a7c57b36e)

- after

![image](https://github.com/user-attachments/assets/f4ba3440-7c1c-4ec1-bc54-4d2ab431eef5)

---------

Signed-off-by: Zakhar Bessarab <z.bessarab@victoriametrics.com>
(cherry picked from commit 837d0d136d)
2024-10-24 12:43:06 -03:00

326 lines
8.4 KiB
Go

package mergeset
import (
"bytes"
"fmt"
"math/rand"
"os"
"sync"
"sync/atomic"
"testing"
)
func TestTableOpenClose(t *testing.T) {
const path = "TestTableOpenClose"
if err := os.RemoveAll(path); err != nil {
t.Fatalf("cannot remove %q: %s", path, err)
}
defer func() {
_ = os.RemoveAll(path)
}()
// Create a new table
var isReadOnly atomic.Bool
tb := MustOpenTable(path, nil, nil, &isReadOnly)
// Close it
tb.MustClose()
// Re-open created table multiple times.
for i := 0; i < 4; i++ {
tb := MustOpenTable(path, nil, nil, &isReadOnly)
tb.MustClose()
}
}
func TestTableAddItemsTooLongItem(t *testing.T) {
const path = "TestTableAddItemsTooLongItem"
if err := os.RemoveAll(path); err != nil {
t.Fatalf("cannot remove %q: %s", path, err)
}
var isReadOnly atomic.Bool
tb := MustOpenTable(path, nil, nil, &isReadOnly)
tb.AddItems([][]byte{make([]byte, maxInmemoryBlockSize+1)})
tb.MustClose()
_ = os.RemoveAll(path)
}
func TestTableAddItemsSerial(t *testing.T) {
r := rand.New(rand.NewSource(1))
const path = "TestTableAddItemsSerial"
if err := os.RemoveAll(path); err != nil {
t.Fatalf("cannot remove %q: %s", path, err)
}
defer func() {
_ = os.RemoveAll(path)
}()
var flushes atomic.Uint64
flushCallback := func() {
flushes.Add(1)
}
var isReadOnly atomic.Bool
tb := MustOpenTable(path, flushCallback, nil, &isReadOnly)
const itemsCount = 10e3
testAddItemsSerial(r, tb, itemsCount)
// Verify items count after pending items flush.
tb.DebugFlush()
if flushes.Load() == 0 {
t.Fatalf("unexpected zero flushes")
}
var m TableMetrics
tb.UpdateMetrics(&m)
if n := m.TotalItemsCount(); n != itemsCount {
t.Fatalf("unexpected itemsCount; got %d; want %v", n, itemsCount)
}
tb.MustClose()
// Re-open the table and make sure itemsCount remains the same.
testReopenTable(t, path, itemsCount)
// Add more items in order to verify merge between inmemory parts and file-based parts.
tb = MustOpenTable(path, nil, nil, &isReadOnly)
const moreItemsCount = itemsCount * 3
testAddItemsSerial(r, tb, moreItemsCount)
tb.MustClose()
// Re-open the table and verify itemsCount again.
testReopenTable(t, path, itemsCount+moreItemsCount)
}
func testAddItemsSerial(r *rand.Rand, tb *Table, itemsCount int) {
for i := 0; i < itemsCount; i++ {
item := getRandomBytes(r)
if len(item) > maxInmemoryBlockSize {
item = item[:maxInmemoryBlockSize]
}
tb.AddItems([][]byte{item})
}
}
func TestTableCreateSnapshotAt(t *testing.T) {
const path = "TestTableCreateSnapshotAt"
if err := os.RemoveAll(path); err != nil {
t.Fatalf("cannot remove %q: %s", path, err)
}
var isReadOnly atomic.Bool
tb := MustOpenTable(path, nil, nil, &isReadOnly)
// Write a lot of items into the table, so background merges would start.
const itemsCount = 3e5
for i := 0; i < itemsCount; i++ {
item := []byte(fmt.Sprintf("item %d", i))
tb.AddItems([][]byte{item})
}
// Close and open the table in order to flush all the data to disk before creating snapshots.
// See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/4272#issuecomment-1550221840
tb.MustClose()
tb = MustOpenTable(path, nil, nil, &isReadOnly)
// Create multiple snapshots.
snapshot1 := path + "-test-snapshot1"
if err := tb.CreateSnapshotAt(snapshot1); err != nil {
t.Fatalf("cannot create snapshot1: %s", err)
}
snapshot2 := path + "-test-snapshot2"
if err := tb.CreateSnapshotAt(snapshot2); err != nil {
t.Fatalf("cannot create snapshot2: %s", err)
}
// Verify snapshots contain all the data.
tb1 := MustOpenTable(snapshot1, nil, nil, &isReadOnly)
tb2 := MustOpenTable(snapshot2, nil, nil, &isReadOnly)
var ts, ts1, ts2 TableSearch
ts.Init(tb, false)
ts1.Init(tb1, false)
ts2.Init(tb2, false)
for i := 0; i < itemsCount; i++ {
key := []byte(fmt.Sprintf("item %d", i))
if err := ts.FirstItemWithPrefix(key); err != nil {
t.Fatalf("cannot find item[%d]=%q in the original table: %s", i, key, err)
}
if !bytes.Equal(key, ts.Item) {
t.Fatalf("unexpected item found for key=%q in the original table; got %q", key, ts.Item)
}
if err := ts1.FirstItemWithPrefix(key); err != nil {
t.Fatalf("cannot find item[%d]=%q in snapshot1: %s", i, key, err)
}
if !bytes.Equal(key, ts1.Item) {
t.Fatalf("unexpected item found for key=%q in snapshot1; got %q", key, ts1.Item)
}
if err := ts2.FirstItemWithPrefix(key); err != nil {
t.Fatalf("cannot find item[%d]=%q in snapshot2: %s", i, key, err)
}
if !bytes.Equal(key, ts2.Item) {
t.Fatalf("unexpected item found for key=%q in snapshot2; got %q", key, ts2.Item)
}
}
ts1.MustClose()
ts2.MustClose()
// Close and remove tables.
tb2.MustClose()
tb1.MustClose()
tb.MustClose()
_ = os.RemoveAll(snapshot2)
_ = os.RemoveAll(snapshot1)
_ = os.RemoveAll(path)
}
func TestTableAddItemsConcurrentStress(t *testing.T) {
const path = "TestTableAddItemsConcurrentStress"
if err := os.RemoveAll(path); err != nil {
t.Fatalf("cannot remove %q: %s", path, err)
}
defer func() {
_ = os.RemoveAll(path)
}()
rawItemsShardsPerTableOrig := rawItemsShardsPerTable
maxBlocksPerShardOrig := maxBlocksPerShard
rawItemsShardsPerTable = 10
maxBlocksPerShard = 3
defer func() {
rawItemsShardsPerTable = rawItemsShardsPerTableOrig
maxBlocksPerShard = maxBlocksPerShardOrig
}()
var flushes atomic.Uint64
flushCallback := func() {
flushes.Add(1)
}
prepareBlock := func(data []byte, items []Item) ([]byte, []Item) {
return data, items
}
blocksNeeded := rawItemsShardsPerTable * maxBlocksPerShard * 10
testAddItems := func(tb *Table) {
itemsBatch := make([][]byte, 0)
for j := 0; j < blocksNeeded; j++ {
item := bytes.Repeat([]byte{byte(j)}, maxInmemoryBlockSize-10)
itemsBatch = append(itemsBatch, item)
}
tb.AddItems(itemsBatch)
}
var isReadOnly atomic.Bool
tb := MustOpenTable(path, flushCallback, prepareBlock, &isReadOnly)
testAddItems(tb)
// Verify items count after pending items flush.
tb.DebugFlush()
if flushes.Load() == 0 {
t.Fatalf("unexpected zero flushes")
}
var m TableMetrics
tb.UpdateMetrics(&m)
if n := m.TotalItemsCount(); n != uint64(blocksNeeded) {
t.Fatalf("unexpected itemsCount; got %d; want %v", n, blocksNeeded)
}
tb.MustClose()
// Re-open the table and make sure itemsCount remains the same.
testReopenTable(t, path, blocksNeeded)
}
func TestTableAddItemsConcurrent(t *testing.T) {
const path = "TestTableAddItemsConcurrent"
if err := os.RemoveAll(path); err != nil {
t.Fatalf("cannot remove %q: %s", path, err)
}
defer func() {
_ = os.RemoveAll(path)
}()
var flushes atomic.Uint64
flushCallback := func() {
flushes.Add(1)
}
prepareBlock := func(data []byte, items []Item) ([]byte, []Item) {
return data, items
}
var isReadOnly atomic.Bool
tb := MustOpenTable(path, flushCallback, prepareBlock, &isReadOnly)
const itemsCount = 10e3
testAddItemsConcurrent(tb, itemsCount)
// Verify items count after pending items flush.
tb.DebugFlush()
if flushes.Load() == 0 {
t.Fatalf("unexpected zero flushes")
}
var m TableMetrics
tb.UpdateMetrics(&m)
if n := m.TotalItemsCount(); n != itemsCount {
t.Fatalf("unexpected itemsCount; got %d; want %v", n, itemsCount)
}
tb.MustClose()
// Re-open the table and make sure itemsCount remains the same.
testReopenTable(t, path, itemsCount)
// Add more items in order to verify merge between inmemory parts and file-based parts.
tb = MustOpenTable(path, nil, nil, &isReadOnly)
const moreItemsCount = itemsCount * 3
testAddItemsConcurrent(tb, moreItemsCount)
tb.MustClose()
// Re-open the table and verify itemsCount again.
testReopenTable(t, path, itemsCount+moreItemsCount)
}
func testAddItemsConcurrent(tb *Table, itemsCount int) {
const goroutinesCount = 6
workCh := make(chan int, itemsCount)
var wg sync.WaitGroup
for i := 0; i < goroutinesCount; i++ {
wg.Add(1)
go func(n int) {
defer wg.Done()
r := rand.New(rand.NewSource(int64(n)))
for range workCh {
item := getRandomBytes(r)
if len(item) > maxInmemoryBlockSize {
item = item[:maxInmemoryBlockSize]
}
tb.AddItems([][]byte{item})
}
}(i)
}
for i := 0; i < itemsCount; i++ {
workCh <- i
}
close(workCh)
wg.Wait()
}
func testReopenTable(t *testing.T, path string, itemsCount int) {
t.Helper()
for i := 0; i < 10; i++ {
var isReadOnly atomic.Bool
tb := MustOpenTable(path, nil, nil, &isReadOnly)
var m TableMetrics
tb.UpdateMetrics(&m)
if n := m.TotalItemsCount(); n != uint64(itemsCount) {
t.Fatalf("unexpected itemsCount after re-opening; got %d; want %v", n, itemsCount)
}
tb.MustClose()
}
}