/chainstore

Lightweight key-value interface to a bunch of storage engines with middleware support, organized as a chain of operations; written in Go

Primary LanguageGo

Chainstore

Chainstore is simple key-value interface to a variety of storage engines organized as a chain of operations. A store adapter is just an engine interface to Open, Close, Put, Get, and Del . Each store has their own inherent properties and so when chained together, it makes for a useful combinations of data caching, flow and persistence depending on your application.

Here is an example of Boltdb and S3 stores chained together to provide fast read/writes to a local working dataset of 500MB and async S3 access for long-term persistence / retrieval. Check out the LRUManager below too, its wrapped around Boltdb to make sure only the least-recently-used key/values are persisted -- the manager can be used with any of the stores and with the chain, which is pretty nifty. This example is also here: example/main.go.

package main

import (
	"fmt"
	"os"
	"time"
	"log"

	"github.com/pressly/chainstore"
	"github.com/pressly/chainstore/boltstore"
	"github.com/pressly/chainstore/lrumgr"
	"github.com/pressly/chainstore/metricsmgr"
	"github.com/pressly/chainstore/s3store"
	"golang.org/x/net/context"
)

var (
	bucketID  string
	accessKey string
	secretKey string
)

func init() {
	bucketID = os.Getenv("S3_BUCKET")
	accessKey = os.Getenv("S3_ACCESS_KEY")
	secretKey = os.Getenv("S3_SECRET_KEY")
}

func main() {
	ctx := context.Background()

	diskStore := lrumgr.New(500*1024*1024, // 500MB of working data
		metricsmgr.New("chainstore.ex.bolt",
			boltstore.New("/tmp/store.db", "myBucket"),
		),
	)

	remoteStore := metricsmgr.New("chainstore.ex.s3",
		// NOTE: you'll have to supply your own keys in order for this example to work properly
		s3store.New(bucketID, accessKey, secretKey),
	)

	dataStore := chainstore.New(diskStore, remoteStore)

	// OR.. define inline. Except, I wanted to show store independence & state.
	/*
		dataStore := chainstore.New(
			lrumgr.New(500*1024*1024, // 500MB of working data
				metricsmgr.New("chainstore.ex.bolt",
					boltstore.New("/tmp/store.db", "myBucket"),
				),
			),
			metricsmgr.New("chainstore.ex.s3",
				// NOTE: you'll have to supply your own keys in order for this example to work properly
				s3store.New("myBucket", "access-key", "secret-key"),
			),
		)
	*/

	var err error

	err = dataStore.Open()
	if err != nil {
		log.Fatalf("Open: %q", err)
	}

	// Since we've used the metricsManager above (metricsmgr), any calls to the boltstore
	// and s3store will be measured. Next is to send metrics to librato, graphite, influxdb,
	// whatever.. via github.com/goware/go-metrics
	// go librato.Librato(metrics.DefaultRegistry, 10e9, ...)

	//--

	// Save the object in the chain. It will be Put() synchronously into diskStore,
	// the boltdb engine, and then immediately dispatch background Put()'s to the
	// other stores down the chain, in this case S3.
	fmt.Println("Example 1...")
	obj := []byte{1, 2, 3}
	err = dataStore.Put(ctx, "k", obj)
	if err != nil {
		log.Fatalf("Put: %q", err)
	}
	fmt.Println("Put 'k':", obj, "in the chain")

	v, err := dataStore.Get(ctx, "k")
	if err != nil {
		log.Fatalf("Put: %q", err)
	}
	fmt.Println("Grabbing 'k' from the chain:", v) // => [1 2 3]

	// For demonstration, let's grab the key directly from the store instead of
	// through the chain. This is pretty much the same as above, as the chain's Get()
	// stops once it finds the object.
	v, err = diskStore.Get(ctx, "k")
	if err != nil {
		log.Fatalf("Put: %q", err)
	}
	fmt.Println("Grabbing 'k' directly from boltdb:", v) // => [1 2 3]

	// lets pause for a moment and then try to retrieve the value from the s3 store
	time.Sleep(1e9)

	// Grab the object from s3
	v, err = remoteStore.Get(ctx, "k")
	if err != nil {
		log.Fatalf("Put: %q", err)
	}
	fmt.Println("Grabbing 'k' directly from s3:", v) // => [1 2 3]

	// Delete the object from everywhere
	dataStore.Del(ctx, "k")
	time.Sleep(1e9) // pause for s3 demo
	v, _ = dataStore.Get(ctx, "k")
	fmt.Println("Deleted 'k' from the chain (all stores). Get(k) returns:", v)

	//--

	// Another interesting behavior of the chain is when doing a Get(), it goes down
	// the entire chain looking for the value, and when found, it will Put() that
	// object back up the chain for subsequent retrievals. Lets see..
	fmt.Println("Example 2...")
	obj = []byte("hope you enjoy")
	err = dataStore.Put(ctx, "hi", obj)
	if err != nil {
		log.Fatalf("Put: %q", err)
	}
	fmt.Println("Put 'hi':", obj, "in the chain")
	time.Sleep(1e9) // lets wait for s3 again with more then enough time

	err = diskStore.Del(ctx, "hi")
	if err != nil {
		log.Fatalf("Get: %q", err)
	}

	v, _ = diskStore.Get(ctx, "hi")
	fmt.Println("Delete 'hi' from boltdb. diskStore.Get(k) returns:", v)

	v, err = dataStore.Get(ctx, "hi")
	if err != nil {
		log.Fatalf("Get: %q", err)
	}
	fmt.Println("Let's ask the chain for 'hi':", v)
	time.Sleep(1e9) // pause for bg routine to fill our local cache

	// The diskStore now has the value again from remoteStore lower down the chain.
	v, err = diskStore.Get(ctx, "hi")
	if err != nil {
		log.Fatalf("Get: %q", err)
	}
	fmt.Println("Now, let's ask our diskStore again! diskStore.Get(k) returns:", v)

	// Also.. even though it hasn't been demonstrated here, the diskStore will only
	// store a max of 500MB (as defined with diskLru) worth of objects. Give it a shot.
}

/* OUTPUT:

Example 1...
Put 'k': [1 2 3] in the chain
Grabbing 'k' from the chain: [1 2 3]
Grabbing 'k' directly from boltdb: [1 2 3]
Grabbing 'k' directly from s3: [1 2 3]
Deleted 'k' from the chain (all stores). Get(k) returns: []
Example 2...
Put 'hi': [104 111 112 101 32 121 111 117 32 101 110 106 111 121] in the chain
Delete 'hi' from boltdb. diskStore.Get(k) returns: []
Let's ask the chain for 'hi': [104 111 112 101 32 121 111 117 32 101 110 106 111 121]
Now, let's ask our diskStore again! diskStore.Get(k) returns: [104 111 112 101 32 121 111 117 32 101 110 106 111 121]

*/

Currently supported stores: memory, filesystem, boltdb, leveldb, s3, a lru manager, and a metrics manager that can be layered ontop. You can chain these together for different behaviours, for example the memstore implementation is just a simple map[string][]byte with the LRU cache manager (lrumgr).

Thx to other great projects:

Changelog

TODO / Ideas

  • Error channel where bad puts are communicated so they can be properly handled further down the chain

  • Idea: provide option to hash the input keys which would make each key fixed-length and smaller footprint everywhere

  • Timeout (with error notification) when adding an item to a store (ie. 60 seconds max to confirm)

  • Consider a 'config' structure to pass to stores that can configure things like: * For s3 store, add ACL with options: private, public_read, public_read_write, authenticated_read

License

Copyright (c) 2014-2015 Peter Kieltyka / Pressly Inc. www.pressly.com

MIT License

Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.