groupcache

A modified version of group cache with support for context.Context, go modules, and explicit key removal taken from mailgun's fork. See the CHANGELOG for a complete list of modifications done by mailgun.

Summary

groupcache is a caching and cache-filling library, intended as a replacement for memcached in many cases.

Modifications from original library

• Replace Mailgun's weird md5 + fnv1 hashing with XXH3

• Use SIEVE patch instead of LRU

• Support for explicit key removal from a group. Remove() requests are first sent to the peer who owns the key, then the remove request is forwarded to every peer in the groupcache. NOTE: This is a best case design since it is possible a temporary network disruption could occur resulting in remove requests never making it their peers. In practice this scenario is very rare and the system remains very consistent. In case of an inconsistency placing a expiration time on your values will ensure the cluster eventually becomes consistent again.

• Now always populating the hotcache. A more complex algorithm is unnecessary when the LRU cache will ensure the most used values remain in the cache. The evict code ensures the hotcache never overcrowds the maincache.

Comparing Groupcache to memcached

Like memcached, groupcache:

• shards by key to select which peer is responsible for that key

Unlike memcached, groupcache:

• does not require running a separate set of servers, thus massively reducing deployment/configuration pain. groupcache is a client library as well as a server. It connects to its own peers.

• comes with a cache filling mechanism. Whereas memcached just says "Sorry, cache miss", often resulting in a thundering herd of database (or whatever) loads from an unbounded number of clients (which has resulted in several fun outages), groupcache coordinates cache fills such that only one load in one process of an entire replicated set of processes populates the cache, then multiplexes the loaded value to all callers.

• does not support versioned values. If key "foo" is value "bar", key "foo" must always be "bar".

Loading process

In a nutshell, a groupcache lookup of Get("foo") looks like:

(On machine #5 of a set of N machines running the same code)

1. Is the value of "foo" in local memory because it's super hot? If so, use it.

2. Is the value of "foo" in local memory because peer #5 (the current peer) is the owner of it? If so, use it.

3. Amongst all the peers in my set of N, am I the owner of the key "foo"? (e.g. does it consistent hash to 5?) If so, load it. If other callers come in, via the same process or via RPC requests from peers, they block waiting for the load to finish and get the same answer. If not, RPC to the peer that's the owner and get the answer. If the RPC fails, just load it locally (still with local dup suppression).

Example

import (
    "context"
    "fmt"
    "log"
    "time"

    "github.com/mailgun/groupcache/v2"
)

func ExampleUsage() {

    // NOTE: It is important to pass the same peer `http://192.168.1.1:8080` to `NewHTTPPoolOpts`
    // which is provided to `pool.Set()` so the pool can identify which of the peers is our instance.
    // The pool will not operate correctly if it can't identify which peer is our instance.
    
    // Pool keeps track of peers in our cluster and identifies which peer owns a key.
    pool := groupcache.NewHTTPPoolOpts("http://192.168.1.1:8080", &groupcache.HTTPPoolOptions{})

    // Add more peers to the cluster You MUST Ensure our instance is included in this list else
    // determining who owns the key accross the cluster will not be consistent, and the pool won't
    // be able to determine if our instance owns the key.
    pool.Set("http://192.168.1.1:8080", "http://192.168.1.2:8080", "http://192.168.1.3:8080")

    server := http.Server{
        Addr:    "192.168.1.1:8080",
        Handler: pool,
    }

    // Start a HTTP server to listen for peer requests from the groupcache
    go func() {
        log.Printf("Serving....\n")
        if err := server.ListenAndServe(); err != nil {
            log.Fatal(err)
        }
    }()
    defer server.Shutdown(context.Background())

    // Create a new group cache with a max cache size of 3MB
    group := groupcache.NewGroup("users", 3000000, groupcache.GetterFunc(
        func(ctx context.Context, id string, dest groupcache.Sink) error {

            // Returns a protobuf struct `User`
            user, err := fetchUserFromMongo(ctx, id)
            if err != nil {
                return err
            }

            // Set the user in the groupcache to expire after 5 minutes
            return dest.SetProto(&user, time.Now().Add(time.Minute*5))
        },
    ))

    var user User

    ctx, cancel := context.WithTimeout(context.Background(), time.Millisecond*500)
    defer cancel()

    if err := group.Get(ctx, "12345", groupcache.ProtoSink(&user)); err != nil {
        log.Fatal(err)
    }

    fmt.Printf("-- User --\n")
    fmt.Printf("Id: %s\n", user.Id)
    fmt.Printf("Name: %s\n", user.Name)
    fmt.Printf("Age: %d\n", user.Age)
    fmt.Printf("IsSuper: %t\n", user.IsSuper)

    // Remove the key from the groupcache
    if err := group.Remove(ctx, "12345"); err != nil {
        log.Fatal(err)
    }
}

Note

The call to groupcache.NewHTTPPoolOpts() is a bit misleading. NewHTTPPoolOpts() creates a new pool internally within the groupcache package where it is uitilized by any groups created. The pool returned is only a pointer to the internallly registered pool so the caller can update the peers in the pool as needed.