What causes "Database redis://%s is not empty" in juicedata/juicefs?

Use an empty Redis DB (fresh instance or an unused DB number) as the load target.

Database redis://%s is not empty

Error message

Database redis://%s is not empty

What it means

Returned by LoadMeta (redis.go:5197) for a non-cluster Redis target: DBSize() reports more than 0 keys, so the destination database is not empty. LoadMeta refuses to proceed because loading a dump into a non-empty DB would produce a corrupted merge of two metadata sets.

Source

Thrown at pkg/meta/redis.go:5197

	tryExec()
}

func (m *redisMeta) LoadMeta(r io.Reader) (err error) {
	ctx := Background()
	if _, ok := m.rdb.(*redis.ClusterClient); ok {
		err = m.scan(ctx, "*", func(keys []string) error {
			return fmt.Errorf("found key with same prefix: %s", keys[0])
		})
		if err != nil {
			return err
		}
	} else {
		dbsize, err := m.rdb.DBSize(ctx).Result()
		if err != nil {
			return err
		}
		if dbsize > 0 {
			return fmt.Errorf("Database redis://%s is not empty", m.addr)
		}
	}

	p := m.rdb.TxPipeline()
	tryExec := func() {
		if p.Len() > 1000 {
			if rs, err := p.Exec(ctx); err != nil {
				for i, r := range rs {
					if r.Err() != nil {
						logger.Errorf("failed command %d %+v: %s", i, r, r.Err())
						break
					}
				}
				panic(err)
			}
		}
	}
	defer func() {

View on GitHub (pinned to c9a67b23e8)

Solutions

Use an empty Redis DB (fresh instance or an unused DB number) as the load target.
If the existing keys are disposable, run FLUSHDB on the target DB and retry.
Verify the Redis URL/DB index in your command points to the intended destination.
If a previous load partially completed, flush and re-load from the original dump to avoid a mixed state.

Example fix

// before
juicefs load redis://shared:6379/0 backup.json   // DB 0 in use
// after
redis-cli -h shared -n 0 flushdb
juicefs load redis://shared:6379/0 backup.json

Defensive patterns

Strategy: validation

Validate before calling

// shell: refuse to load into a non-empty DB
dbsize=$(redis-cli -h target -p 6379 -n 0 dbsize)
[ "$dbsize" = "0" ] || { echo "DB not empty ($dbsize keys)"; exit 1; }

Try / catch

if err := meta.LoadMeta(ctx, r); err != nil {
    if strings.Contains(err.Error(), "is not empty") {
        // pick an empty DB index or FLUSHDB the target, then retry
    }
    return err
}

Prevention

Load into a dedicated, empty Redis instance or unused DB index.
FLUSHDB the target before every restore in scripts.
Never point `juicefs load` at a live volume's metadata URL.
After a failed load, flush before retrying — partial keys remain.

When it happens

Trigger: Running `juicefs load` against a standalone Redis URL whose selected DB already contains keys (e.g. an existing volume's metadata, or leftover keys from a previous load).

Common situations: Restoring over a live volume; loading into DB 0 of a shared Redis server; a failed/partial previous load leaving keys behind.

Understand the failure class

Background: "already exists" / EEXIST / FileAlreadyExistsException: what the 'file already exists' error means and how to fix it — this error's family across 37 libraries.

Related errors

AI-assisted analysis of juicedata/juicefs@c9a67b23e8 (2026-09-06). Data as JSON: /api/errors/530f8a3483782cdf. Report an issue: GitHub.