mirror of
https://github.com/VictoriaMetrics/VictoriaMetrics.git
synced 2024-11-23 20:37:12 +01:00
837d0d136d
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>
168 lines
3.9 KiB
Go
168 lines
3.9 KiB
Go
package mergeset
|
|
|
|
import (
|
|
"fmt"
|
|
"math/rand"
|
|
"os"
|
|
"sort"
|
|
"sync/atomic"
|
|
"testing"
|
|
"time"
|
|
)
|
|
|
|
func TestMain(m *testing.M) {
|
|
n := m.Run()
|
|
os.Exit(n)
|
|
}
|
|
|
|
func TestTableSearchSerial(t *testing.T) {
|
|
const path = "TestTableSearchSerial"
|
|
if err := os.RemoveAll(path); err != nil {
|
|
t.Fatalf("cannot remove %q: %s", path, err)
|
|
}
|
|
defer func() {
|
|
_ = os.RemoveAll(path)
|
|
}()
|
|
|
|
const itemsCount = 1e5
|
|
|
|
items := func() []string {
|
|
r := rand.New(rand.NewSource(1))
|
|
tb, items, err := newTestTable(r, path, itemsCount)
|
|
if err != nil {
|
|
t.Fatalf("cannot create test table: %s", err)
|
|
}
|
|
defer tb.MustClose()
|
|
if err := testTableSearchSerial(tb, items); err != nil {
|
|
t.Fatalf("unexpected error: %s", err)
|
|
}
|
|
return items
|
|
}()
|
|
|
|
func() {
|
|
// Re-open the table and verify the search works.
|
|
var isReadOnly atomic.Bool
|
|
tb := MustOpenTable(path, nil, nil, &isReadOnly)
|
|
defer tb.MustClose()
|
|
if err := testTableSearchSerial(tb, items); err != nil {
|
|
t.Fatalf("unexpected error: %s", err)
|
|
}
|
|
}()
|
|
}
|
|
|
|
func TestTableSearchConcurrent(t *testing.T) {
|
|
const path = "TestTableSearchConcurrent"
|
|
if err := os.RemoveAll(path); err != nil {
|
|
t.Fatalf("cannot remove %q: %s", path, err)
|
|
}
|
|
defer func() {
|
|
_ = os.RemoveAll(path)
|
|
}()
|
|
|
|
const itemsCount = 1e5
|
|
items := func() []string {
|
|
r := rand.New(rand.NewSource(2))
|
|
tb, items, err := newTestTable(r, path, itemsCount)
|
|
if err != nil {
|
|
t.Fatalf("cannot create test table: %s", err)
|
|
}
|
|
defer tb.MustClose()
|
|
if err := testTableSearchConcurrent(tb, items); err != nil {
|
|
t.Fatalf("unexpected error: %s", err)
|
|
}
|
|
return items
|
|
}()
|
|
|
|
// Re-open the table and verify the search works.
|
|
func() {
|
|
var isReadOnly atomic.Bool
|
|
tb := MustOpenTable(path, nil, nil, &isReadOnly)
|
|
defer tb.MustClose()
|
|
if err := testTableSearchConcurrent(tb, items); err != nil {
|
|
t.Fatalf("unexpected error: %s", err)
|
|
}
|
|
}()
|
|
}
|
|
|
|
func testTableSearchConcurrent(tb *Table, items []string) error {
|
|
const goroutines = 5
|
|
ch := make(chan error, goroutines)
|
|
for i := 0; i < goroutines; i++ {
|
|
go func() {
|
|
ch <- testTableSearchSerial(tb, items)
|
|
}()
|
|
}
|
|
for i := 0; i < goroutines; i++ {
|
|
select {
|
|
case err := <-ch:
|
|
if err != nil {
|
|
return fmt.Errorf("unexpected error: %w", err)
|
|
}
|
|
case <-time.After(time.Second * 5):
|
|
return fmt.Errorf("timeout")
|
|
}
|
|
}
|
|
return nil
|
|
}
|
|
|
|
func testTableSearchSerial(tb *Table, items []string) error {
|
|
var ts TableSearch
|
|
ts.Init(tb, false)
|
|
for _, key := range []string{
|
|
"",
|
|
"123",
|
|
"9",
|
|
"892",
|
|
"2384329",
|
|
"fdsjflfdf",
|
|
items[0],
|
|
items[len(items)-1],
|
|
items[len(items)/2],
|
|
} {
|
|
n := sort.Search(len(items), func(i int) bool {
|
|
return key <= items[i]
|
|
})
|
|
ts.Seek([]byte(key))
|
|
for n < len(items) {
|
|
item := items[n]
|
|
if !ts.NextItem() {
|
|
return fmt.Errorf("missing item %q at position %d when searching for %q", item, n, key)
|
|
}
|
|
if string(ts.Item) != item {
|
|
return fmt.Errorf("unexpected item found at position %d when searching for %q; got %q; want %q", n, key, ts.Item, item)
|
|
}
|
|
n++
|
|
}
|
|
if ts.NextItem() {
|
|
return fmt.Errorf("superfluous item found at position %d when searching for %q: %q", n, key, ts.Item)
|
|
}
|
|
if err := ts.Error(); err != nil {
|
|
return fmt.Errorf("unexpected error when searching for %q: %w", key, err)
|
|
}
|
|
}
|
|
ts.MustClose()
|
|
return nil
|
|
}
|
|
|
|
func newTestTable(r *rand.Rand, path string, itemsCount int) (*Table, []string, error) {
|
|
var flushes atomic.Uint64
|
|
flushCallback := func() {
|
|
flushes.Add(1)
|
|
}
|
|
var isReadOnly atomic.Bool
|
|
tb := MustOpenTable(path, flushCallback, nil, &isReadOnly)
|
|
items := make([]string, itemsCount)
|
|
for i := 0; i < itemsCount; i++ {
|
|
item := fmt.Sprintf("%d:%d", r.Intn(1e9), i)
|
|
tb.AddItems([][]byte{[]byte(item)})
|
|
items[i] = item
|
|
}
|
|
tb.DebugFlush()
|
|
if itemsCount > 0 && flushes.Load() == 0 {
|
|
return nil, nil, fmt.Errorf("unexpeted zero flushes for itemsCount=%d", itemsCount)
|
|
}
|
|
|
|
sort.Strings(items)
|
|
return tb, items, nil
|
|
}
|