GC doesn't seem to run

[lkp-k]

What version of Badger are you using?

Latest v4

What version of Go are you using?

1.20.3

Have you tried reproducing the issue with the latest release?

Yes

What is the hardware spec (RAM, CPU, OS)?

16gb ram, i5 intel, mac os

What steps will reproduce the bug?

package main

import (
	"fmt"
	"log"
	"math/rand"
	"strconv"
	"time"

	"github.com/dgraph-io/badger/v4"
)

var db *badger.DB

const charset = "abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ0123456789"

func generateRandomString(length int) string {
	b := make([]byte, length)
	for i := range b {
		b[i] = charset[rand.Intn(len(charset))]
	}
	return string(b)
}

func main() {
	opts := badger.DefaultOptions("badger")
	opts.ValueLogFileSize = 1 << 20
	var err error
	db, err = badger.Open(opts)
	if err != nil {
		log.Fatal(err)
	}
	defer db.Close()

	fmt.Println("Start inserting 5 million items.")

	for i := 0; i < 5_000_000; i++ {
		err := db.Update(func(txn *badger.Txn) error {
			key := strconv.Itoa(i)
			value := []byte(generateRandomString(300))

			e := badger.NewEntry([]byte(key), value).WithTTL(30 * time.Second)

			err := txn.SetEntry(e)
			if err != nil {
				return err
			}
			return nil
		})

		if err != nil {
			log.Fatal("Error while inserting:", err)
		}
	}

	fmt.Println("Successfully inserted 5 million items.")

	startGarbageCollection()
}

func startGarbageCollection() {
	ticker := time.NewTicker(10 * time.Second)
	defer ticker.Stop()
	for range ticker.C {
	again:
		err := db.RunValueLogGC(0.01)
		fmt.Println(err)
		if err == nil {
			goto again
		}
	}
}

Expected behavior and actual result.

Since the keys are expiring in 30 seconds, I thought that the files in badger/ would get removed.

However, heres what I see:

Additional information

No response

[mangalaman93]

may be related to #1995. Thanks for filling an issue with code to reproduce. I will try to find some time to look into it.

[wangyang0918]

Maybe the compaction is not triggered. Similar to https://discuss.dgraph.io/t/gc-may-not-work-in-some-cases/17197

[lkp-k]

Hi Friends. Any updates on this? Ty

[GraphR00t]

@mangalaman93 Would it possible for you to take a look when you have some time ? Thank you !

[github-actions]

This issue has been stale for 60 days and will be closed automatically in 7 days. Comment to keep it open.

[GraphR00t]

Any updates ? Thanks !

[chrisDeFouRire]

Here's some information I hope could be of interest:

• just running the ValueLogGC will not always trigger a GC
• there must be compaction before ValueLogGC clears value log files
• that's me testing with a 2 min TTL
• I modified the LSM params to cause frequent compactions, then I saw the GC running... sometimes !
• if no key was cleared by compaction, no value will be cleared of ValueLog

So if you want GC to happen more often, you must make it so compaction happens more often.

In my test, I used the following (extreme!) params before I could see things happen over a few minutes (with constant additions... the LSM doesn't fill in fast with high threshold).

badger.Open(badger.DefaultOptions("archiver").
WithValueLogMaxEntries(1000).WithValueThreshold(100).WithCompactL0OnClose(true).
WithBaseLevelSize(10000).WithNumLevelZeroTables(2).WithBaseTableSize(4000).WithMemTableSize(1000000))

Of course I do not recommend these params, I just wanted to make sure Badger does indeed GC, with the right conditions. Make sure you understand each param and don't blame me 🤣

All in all, GC works, if compaction occurs and compacts TTL'ed entries.