open-nomad/vendor/github.com/boltdb/bolt

Bolt

Bolt is a pure Go key/value store inspired by Howard Chu's LMDB project. The goal of the project is to provide a simple, fast, and reliable database for projects that don't require a full database server such as Postgres or MySQL.

Since Bolt is meant to be used as such a low-level piece of functionality, simplicity is key. The API will be small and only focus on getting values and setting values. That's it.

Project Status

Bolt is stable and the API is fixed. Full unit test coverage and randomized black box testing are used to ensure database consistency and thread safety. Bolt is currently in high-load production environments serving databases as large as 1TB. Many companies such as Shopify and Heroku use Bolt-backed services every day.

Getting Started

Installing

To start using Bolt, install Go and run go get:

$ go get github.com/boltdb/bolt/...

This will retrieve the library and install the bolt command line utility into your $GOBIN path.

Opening a database

The top-level object in Bolt is a DB. It is represented as a single file on your disk and represents a consistent snapshot of your data.

To open your database, simply use the bolt.Open() function:

package main

import (
	"log"

	"github.com/boltdb/bolt"
)

func main() {
	// Open the my.db data file in your current directory.
	// It will be created if it doesn't exist.
	db, err := bolt.Open("my.db", 0600, nil)
	if err != nil {
		log.Fatal(err)
	}
	defer db.Close()

	...
}

Please note that Bolt obtains a file lock on the data file so multiple processes cannot open the same database at the same time. Opening an already open Bolt database will cause it to hang until the other process closes it. To prevent an indefinite wait you can pass a timeout option to the Open() function:

db, err := bolt.Open("my.db", 0600, &bolt.Options{Timeout: 1 * time.Second})

Transactions

Bolt allows only one read-write transaction at a time but allows as many read-only transactions as you want at a time. Each transaction has a consistent view of the data as it existed when the transaction started.

Individual transactions and all objects created from them (e.g. buckets, keys) are not thread safe. To work with data in multiple goroutines you must start a transaction for each one or use locking to ensure only one goroutine accesses a transaction at a time. Creating transaction from the DB is thread safe.

Read-only transactions and read-write transactions should not depend on one another and generally shouldn't be opened simultaneously in the same goroutine. This can cause a deadlock as the read-write transaction needs to periodically re-map the data file but it cannot do so while a read-only transaction is open.

Read-write transactions

To start a read-write transaction, you can use the DB.Update() function:

err := db.Update(func(tx *bolt.Tx) error {
	...
	return nil
})

Inside the closure, you have a consistent view of the database. You commit the transaction by returning nil at the end. You can also rollback the transaction at any point by returning an error. All database operations are allowed inside a read-write transaction.

Always check the return error as it will report any disk failures that can cause your transaction to not complete. If you return an error within your closure it will be passed through.

Read-only transactions

To start a read-only transaction, you can use the DB.View() function:

err := db.View(func(tx *bolt.Tx) error {
	...
	return nil
})

You also get a consistent view of the database within this closure, however, no mutating operations are allowed within a read-only transaction. You can only retrieve buckets, retrieve values, and copy the database within a read-only transaction.

Batch read-write transactions

Each DB.Update() waits for disk to commit the writes. This overhead can be minimized by combining multiple updates with the DB.Batch() function:

err := db.Batch(func(tx *bolt.Tx) error {
	...
	return nil
})

Concurrent Batch calls are opportunistically combined into larger transactions. Batch is only useful when there are multiple goroutines calling it.

The trade-off is that Batch can call the given function multiple times, if parts of the transaction fail. The function must be idempotent and side effects must take effect only after a successful return from DB.Batch().

For example: don't display messages from inside the function, instead set variables in the enclosing scope:

var id uint64
err := db.Batch(func(tx *bolt.Tx) error {
	// Find last key in bucket, decode as bigendian uint64, increment
	// by one, encode back to []byte, and add new key.
	...
	id = newValue
	return nil
})
if err != nil {
	return ...
}
fmt.Println("Allocated ID %d", id)

Managing transactions manually

The DB.View() and DB.Update() functions are wrappers around the DB.Begin() function. These helper functions will start the transaction, execute a function, and then safely close your transaction if an error is returned. This is the recommended way to use Bolt transactions.

However, sometimes you may want to manually start and end your transactions. You can use the Tx.Begin() function directly but please be sure to close the transaction.

// Start a writable transaction.
tx, err := db.Begin(true)
if err != nil {
    return err
}
defer tx.Rollback()

// Use the transaction...
_, err := tx.CreateBucket([]byte("MyBucket"))
if err != nil {
    return err
}

// Commit the transaction and check for error.
if err := tx.Commit(); err != nil {
    return err
}

The first argument to DB.Begin() is a boolean stating if the transaction should be writable.

Using buckets

Buckets are collections of key/value pairs within the database. All keys in a bucket must be unique. You can create a bucket using the DB.CreateBucket() function:

db.Update(func(tx *bolt.Tx) error {
	b, err := tx.CreateBucket([]byte("MyBucket"))
	if err != nil {
		return fmt.Errorf("create bucket: %s", err)
	}
	return nil
})

You can also create a bucket only if it doesn't exist by using the Tx.CreateBucketIfNotExists() function. It's a common pattern to call this function for all your top-level buckets after you open your database so you can guarantee that they exist for future transactions.

To delete a bucket, simply call the Tx.DeleteBucket() function.

Using key/value pairs

To save a key/value pair to a bucket, use the Bucket.Put() function:

db.Update(func(tx *bolt.Tx) error {
	b := tx.Bucket([]byte("MyBucket"))
	err := b.Put([]byte("answer"), []byte("42"))
	return err
})

This will set the value of the "answer" key to "42" in the MyBucket bucket. To retrieve this value, we can use the Bucket.Get() function:

db.View(func(tx *bolt.Tx) error {
	b := tx.Bucket([]byte("MyBucket"))
	v := b.Get([]byte("answer"))
	fmt.Printf("The answer is: %s\n", v)
	return nil
})

The Get() function does not return an error because its operation is guaranteed to work (unless there is some kind of system failure). If the key exists then it will return its byte slice value. If it doesn't exist then it will return nil. It's important to note that you can have a zero-length value set to a key which is different than the key not existing.

Use the Bucket.Delete() function to delete a key from the bucket.

Please note that values returned from Get() are only valid while the transaction is open. If you need to use a value outside of the transaction then you must use copy() to copy it to another byte slice.

Autoincrementing integer for the bucket

By using the NextSequence() function, you can let Bolt determine a sequence which can be used as the unique identifier for your key/value pairs. See the example below.

// CreateUser saves u to the store. The new user ID is set on u once the data is persisted.
func (s *Store) CreateUser(u *User) error {
    return s.db.Update(func(tx *bolt.Tx) error {
        // Retrieve the users bucket.
        // This should be created when the DB is first opened.
        b := tx.Bucket([]byte("users"))

        // Generate ID for the user.
        // This returns an error only if the Tx is closed or not writeable.
        // That can't happen in an Update() call so I ignore the error check.
        id, _ = b.NextSequence()
        u.ID = int(id)

        // Marshal user data into bytes.
        buf, err := json.Marshal(u)
        if err != nil {
            return err
        }

        // Persist bytes to users bucket.
        return b.Put(itob(u.ID), buf)
    })
}

// itob returns an 8-byte big endian representation of v.
func itob(v int) []byte {
    b := make([]byte, 8)
    binary.BigEndian.PutUint64(b, uint64(v))
    return b
}

type User struct {
    ID int
    ...
}

Iterating over keys

Bolt stores its keys in byte-sorted order within a bucket. This makes sequential iteration over these keys extremely fast. To iterate over keys we'll use a Cursor:

db.View(func(tx *bolt.Tx) error {
	b := tx.Bucket([]byte("MyBucket"))
	c := b.Cursor()

	for k, v := c.First(); k != nil; k, v = c.Next() {
		fmt.Printf("key=%s, value=%s\n", k, v)
	}

	return nil
})

The cursor allows you to move to a specific point in the list of keys and move forward or backward through the keys one at a time.

The following functions are available on the cursor:

First()  Move to the first key.
Last()   Move to the last key.
Seek()   Move to a specific key.
Next()   Move to the next key.
Prev()   Move to the previous key.

Each of those functions has a return signature of (key []byte, value []byte). When you have iterated to the end of the cursor then Next() will return a nil key. You must seek to a position using First(), Last(), or Seek() before calling Next() or Prev(). If you do not seek to a position then these functions will return a nil key.

During iteration, if the key is non-nil but the value is nil, that means the key refers to a bucket rather than a value. Use Bucket.Bucket() to access the sub-bucket.

Prefix scans

To iterate over a key prefix, you can combine Seek() and bytes.HasPrefix():

db.View(func(tx *bolt.Tx) error {
	c := tx.Bucket([]byte("MyBucket")).Cursor()

	prefix := []byte("1234")
	for k, v := c.Seek(prefix); bytes.HasPrefix(k, prefix); k, v = c.Next() {
		fmt.Printf("key=%s, value=%s\n", k, v)
	}

	return nil
})

Range scans

Another common use case is scanning over a range such as a time range. If you use a sortable time encoding such as RFC3339 then you can query a specific date range like this:

db.View(func(tx *bolt.Tx) error {
	// Assume our events bucket has RFC3339 encoded time keys.
	c := tx.Bucket([]byte("Events")).Cursor()

	// Our time range spans the 90's decade.
	min := []byte("1990-01-01T00:00:00Z")
	max := []byte("2000-01-01T00:00:00Z")

	// Iterate over the 90's.
	for k, v := c.Seek(min); k != nil && bytes.Compare(k, max) <= 0; k, v = c.Next() {
		fmt.Printf("%s: %s\n", k, v)
	}

	return nil
})

ForEach()

You can also use the function ForEach() if you know you'll be iterating over all the keys in a bucket:

db.View(func(tx *bolt.Tx) error {
	b := tx.Bucket([]byte("MyBucket"))
	b.ForEach(func(k, v []byte) error {
		fmt.Printf("key=%s, value=%s\n", k, v)
		return nil
	})
	return nil
})

Nested buckets

You can also store a bucket in a key to create nested buckets. The API is the same as the bucket management API on the DB object:

func (*Bucket) CreateBucket(key []byte) (*Bucket, error)
func (*Bucket) CreateBucketIfNotExists(key []byte) (*Bucket, error)
func (*Bucket) DeleteBucket(key []byte) error

Database backups

Bolt is a single file so it's easy to backup. You can use the Tx.WriteTo() function to write a consistent view of the database to a writer. If you call this from a read-only transaction, it will perform a hot backup and not block your other database reads and writes.

By default, it will use a regular file handle which will utilize the operating system's page cache. See the Tx documentation for information about optimizing for larger-than-RAM datasets.

One common use case is to backup over HTTP so you can use tools like cURL to do database backups:

func BackupHandleFunc(w http.ResponseWriter, req *http.Request) {
	err := db.View(func(tx *bolt.Tx) error {
		w.Header().Set("Content-Type", "application/octet-stream")
		w.Header().Set("Content-Disposition", `attachment; filename="my.db"`)
		w.Header().Set("Content-Length", strconv.Itoa(int(tx.Size())))
		_, err := tx.WriteTo(w)
		return err
	})
	if err != nil {
		http.Error(w, err.Error(), http.StatusInternalServerError)
	}
}

Then you can backup using this command:

$ curl http://localhost/backup > my.db

Or you can open your browser to http://localhost/backup and it will download automatically.

If you want to backup to another file you can use the Tx.CopyFile() helper function.

Statistics

The database keeps a running count of many of the internal operations it performs so you can better understand what's going on. By grabbing a snapshot of these stats at two points in time we can see what operations were performed in that time range.

For example, we could start a goroutine to log stats every 10 seconds:

go func() {
	// Grab the initial stats.
	prev := db.Stats()

	for {
		// Wait for 10s.
		time.Sleep(10 * time.Second)

		// Grab the current stats and diff them.
		stats := db.Stats()
		diff := stats.Sub(&prev)

		// Encode stats to JSON and print to STDERR.
		json.NewEncoder(os.Stderr).Encode(diff)

		// Save stats for the next loop.
		prev = stats
	}
}()

It's also useful to pipe these stats to a service such as statsd for monitoring or to provide an HTTP endpoint that will perform a fixed-length sample.

Read-Only Mode

Sometimes it is useful to create a shared, read-only Bolt database. To this, set the Options.ReadOnly flag when opening your database. Read-only mode uses a shared lock to allow multiple processes to read from the database but it will block any processes from opening the database in read-write mode.

db, err := bolt.Open("my.db", 0666, &bolt.Options{ReadOnly: true})
if err != nil {
	log.Fatal(err)
}

Resources

For more information on getting started with Bolt, check out the following articles:

• Intro to BoltDB: Painless Performant Persistence by Nate Finch.
• Bolt -- an embedded key/value database for Go by Progville

Comparison with other databases

Postgres, MySQL, & other relational databases

Relational databases structure data into rows and are only accessible through the use of SQL. This approach provides flexibility in how you store and query your data but also incurs overhead in parsing and planning SQL statements. Bolt accesses all data by a byte slice key. This makes Bolt fast to read and write data by key but provides no built-in support for joining values together.

Most relational databases (with the exception of SQLite) are standalone servers that run separately from your application. This gives your systems flexibility to connect multiple application servers to a single database server but also adds overhead in serializing and transporting data over the network. Bolt runs as a library included in your application so all data access has to go through your application's process. This brings data closer to your application but limits multi-process access to the data.

LevelDB, RocksDB

LevelDB and its derivatives (RocksDB, HyperLevelDB) are similar to Bolt in that they are libraries bundled into the application, however, their underlying structure is a log-structured merge-tree (LSM tree). An LSM tree optimizes random writes by using a write ahead log and multi-tiered, sorted files called SSTables. Bolt uses a B+tree internally and only a single file. Both approaches have trade-offs.

If you require a high random write throughput (>10,000 w/sec) or you need to use spinning disks then LevelDB could be a good choice. If your application is read-heavy or does a lot of range scans then Bolt could be a good choice.

One other important consideration is that LevelDB does not have transactions. It supports batch writing of key/values pairs and it supports read snapshots but it will not give you the ability to do a compare-and-swap operation safely. Bolt supports fully serializable ACID transactions.

LMDB

Bolt was originally a port of LMDB so it is architecturally similar. Both use a B+tree, have ACID semantics with fully serializable transactions, and support lock-free MVCC using a single writer and multiple readers.

The two projects have somewhat diverged. LMDB heavily focuses on raw performance while Bolt has focused on simplicity and ease of use. For example, LMDB allows several unsafe actions such as direct writes for the sake of performance. Bolt opts to disallow actions which can leave the database in a corrupted state. The only exception to this in Bolt is DB.NoSync.

There are also a few differences in API. LMDB requires a maximum mmap size when opening an mdb_env whereas Bolt will handle incremental mmap resizing automatically. LMDB overloads the getter and setter functions with multiple flags whereas Bolt splits these specialized cases into their own functions.

Caveats & Limitations

It's important to pick the right tool for the job and Bolt is no exception. Here are a few things to note when evaluating and using Bolt:

• Bolt is good for read intensive workloads. Sequential write performance is also fast but random writes can be slow. You can use DB.Batch() or add a write-ahead log to help mitigate this issue.

• Bolt uses a B+tree internally so there can be a lot of random page access. SSDs provide a significant performance boost over spinning disks.

• Try to avoid long running read transactions. Bolt uses copy-on-write so old pages cannot be reclaimed while an old transaction is using them.

• Byte slices returned from Bolt are only valid during a transaction. Once the transaction has been committed or rolled back then the memory they point to can be reused by a new page or can be unmapped from virtual memory and you'll see an unexpected fault address panic when accessing it.

• Be careful when using Bucket.FillPercent. Setting a high fill percent for buckets that have random inserts will cause your database to have very poor page utilization.

• Use larger buckets in general. Smaller buckets causes poor page utilization once they become larger than the page size (typically 4KB).

• Bulk loading a lot of random writes into a new bucket can be slow as the page will not split until the transaction is committed. Randomly inserting more than 100,000 key/value pairs into a single new bucket in a single transaction is not advised.

• Bolt uses a memory-mapped file so the underlying operating system handles the caching of the data. Typically, the OS will cache as much of the file as it can in memory and will release memory as needed to other processes. This means that Bolt can show very high memory usage when working with large databases. However, this is expected and the OS will release memory as needed. Bolt can handle databases much larger than the available physical RAM, provided its memory-map fits in the process virtual address space. It may be problematic on 32-bits systems.

• The data structures in the Bolt database are memory mapped so the data file will be endian specific. This means that you cannot copy a Bolt file from a little endian machine to a big endian machine and have it work. For most users this is not a concern since most modern CPUs are little endian.

• Because of the way pages are laid out on disk, Bolt cannot truncate data files and return free pages back to the disk. Instead, Bolt maintains a free list of unused pages within its data file. These free pages can be reused by later transactions. This works well for many use cases as databases generally tend to grow. However, it's important to note that deleting large chunks of data will not allow you to reclaim that space on disk.

  For more information on page allocation, see this comment.

Reading the Source

Bolt is a relatively small code base (<3KLOC) for an embedded, serializable, transactional key/value database so it can be a good starting point for people interested in how databases work.

The best places to start are the main entry points into Bolt:

• Open() - Initializes the reference to the database. It's responsible for creating the database if it doesn't exist, obtaining an exclusive lock on the file, reading the meta pages, & memory-mapping the file.

• DB.Begin() - Starts a read-only or read-write transaction depending on the value of the writable argument. This requires briefly obtaining the "meta" lock to keep track of open transactions. Only one read-write transaction can exist at a time so the "rwlock" is acquired during the life of a read-write transaction.

• Bucket.Put() - Writes a key/value pair into a bucket. After validating the arguments, a cursor is used to traverse the B+tree to the page and position where they key & value will be written. Once the position is found, the bucket materializes the underlying page and the page's parent pages into memory as "nodes". These nodes are where mutations occur during read-write transactions. These changes get flushed to disk during commit.

• Bucket.Get() - Retrieves a key/value pair from a bucket. This uses a cursor to move to the page & position of a key/value pair. During a read-only transaction, the key and value data is returned as a direct reference to the underlying mmap file so there's no allocation overhead. For read-write transactions, this data may reference the mmap file or one of the in-memory node values.

• Cursor - This object is simply for traversing the B+tree of on-disk pages or in-memory nodes. It can seek to a specific key, move to the first or last value, or it can move forward or backward. The cursor handles the movement up and down the B+tree transparently to the end user.

• Tx.Commit() - Converts the in-memory dirty nodes and the list of free pages into pages to be written to disk. Writing to disk then occurs in two phases. First, the dirty pages are written to disk and an fsync() occurs. Second, a new meta page with an incremented transaction ID is written and another fsync() occurs. This two phase write ensures that partially written data pages are ignored in the event of a crash since the meta page pointing to them is never written. Partially written meta pages are invalidated because they are written with a checksum.

If you have additional notes that could be helpful for others, please submit them via pull request.

Other Projects Using Bolt

Below is a list of public, open source projects that use Bolt:

• Operation Go: A Routine Mission - An online programming game for Golang using Bolt for user accounts and a leaderboard.
• Bazil - A file system that lets your data reside where it is most convenient for it to reside.
• DVID - Added Bolt as optional storage engine and testing it against Basho-tuned leveldb.
• Skybox Analytics - A standalone funnel analysis tool for web analytics.
• Scuttlebutt - Uses Bolt to store and process all Twitter mentions of GitHub projects.
• Wiki - A tiny wiki using Goji, BoltDB and Blackfriday.
• ChainStore - Simple key-value interface to a variety of storage engines organized as a chain of operations.
• MetricBase - Single-binary version of Graphite.
• Gitchain - Decentralized, peer-to-peer Git repositories aka "Git meets Bitcoin".
• event-shuttle - A Unix system service to collect and reliably deliver messages to Kafka.
• ipxed - Web interface and api for ipxed.
• BoltStore - Session store using Bolt.
• photosite/session - Sessions for a photo viewing site.
• LedisDB - A high performance NoSQL, using Bolt as optional storage.
• ipLocator - A fast ip-geo-location-server using bolt with bloom filters.
• cayley - Cayley is an open-source graph database using Bolt as optional backend.
• bleve - A pure Go search engine similar to ElasticSearch that uses Bolt as the default storage backend.
• tentacool - REST api server to manage system stuff (IP, DNS, Gateway...) on a linux server.
• SkyDB - Behavioral analytics database.
• Seaweed File System - Highly scalable distributed key~file system with O(1) disk read.
• InfluxDB - Scalable datastore for metrics, events, and real-time analytics.
• Freehold - An open, secure, and lightweight platform for your files and data.
• Prometheus Annotation Server - Annotation server for PromDash & Prometheus service monitoring system.
• Consul - Consul is service discovery and configuration made easy. Distributed, highly available, and datacenter-aware.
• Kala - Kala is a modern job scheduler optimized to run on a single node. It is persistent, JSON over HTTP API, ISO 8601 duration notation, and dependent jobs.
• drive - drive is an unofficial Google Drive command line client for *NIX operating systems.
• stow - a persistence manager for objects backed by boltdb.
• buckets - a bolt wrapper streamlining simple tx and key scans.

If you are using Bolt in a project please send a pull request to add it to the list.