lru-cache-for-clusters-as-promised

LRU Cache for Clusters as Promised provides a cluster-safe lru-cache via Promises. For environments not using cluster, the class will provide a Promisified interface to a standard lru-cache.

Each time you call cluster.fork(), a new thread is spawned to run your application. When using a load balancer even if a user is assigned a particular IP and port these values are shared between the workers in your cluster, which means there is no guarantee that the user will use the same workers between requests. Caching the same objects in multiple threads is not an efficient use of memory.

LRU Cache for Clusters as Promised stores a single lru-cache on the master thread which is accessed by the workers via IPC messages. The same lru-cache is shared between workers having a common master, so no memory is wasted.

When creating a new instance and cluster.isMaster === true the shared cache is checked based on the and the shared cache is populated, it will be used instead but acted on locally rather than via IPC messages. If the shared cache is not populated a new LRUCache instance is returned.

install

1npm install --save lru-cache-for-clusters-as-promised

1yarn add lru-cache-for-clusters-as-promised

options

• namespace: string, default "default";
  • The namespace for this cache on the master thread as it is not aware of the worker instances.
• timeout: integer, default 100.
  • The amount of time in milliseconds that a worker will wait for a response from the master before rejecting the Promise.
• failsafe: string, default resolve.
  • When a request times out the Promise will return resolve(undefined) by default, or with a value of reject the return will be reject(Error).
• max: number
  • The maximum items that can be stored in the cache
• maxAge: milliseconds
  • The maximum age for an item to be considered valid
• stale: true|false
  • When true expired items are return before they are removed rather than undefined
• prune: false|crontime string, defaults to false
  • Use a cron job on the master thread to call prune() on your cache at regular intervals specified in "crontime", for example "*/30 * * * * *" would prune the cache every 30 seconds (See node-cron patterns for more info). Also works in single threaded environments not using the cluster module. Passing false to an existing namespace will disable any jobs that are scheduled.
• parse: function, defaults to JSON.parse
  • Pass in a custom parser function to use for deserializing data sent to/from the cache. This is set on the LRUCacheForClustersAsPromised instance and in theory could be different per worker.
• stringify: function, defaults to JSON.stringify
  • Pass in a custom stringifier function to for creating a serializing data sent to/from the cache.

! note that length and dispose are missing as it is not possible to pass functions via IPC messages.

api

static functions

• init(): void
  • Should be called when cluster.isMaster === true to initialize the caches.
• getInstance(options): Promise<LRUCacheForClustersAsPromised>
  • Asynchronously returns an LRUCacheForClustersAsPromised instance once the underlying LRUCache is guaranteed to exist. Uses the same options you would pass to the constructor. When constructed synchronously other methods will ensure the underlying cache is created, but this method can be useful from the worker when you plan to interact with the caches directly. Note that this will slow down the construction time on the worker by a few milliseconds while the cache creation is confirmed.
• getAllCaches(): { key : LRUCache }
  • Synchronously returns a dictionary of the underlying LRUCache caches keyed by namespace. Accessible only when cluster.isMaster === true, otherwise throws an exception.

instance functions

• getCache(): LRUCache
  • Gets the underlying LRUCache. Accessible only when cluster.isMaster === true, otherwise throws an exception.
• set(key, value, maxAge): Promise<void>
  • Sets a value for a key. Specifying the maxAge will cause the value to expire per the stale value or when pruned.
• setObject async (key, object, maxAge): Promise<void>
  • Sets a cache value where the value is an object. Passes the values through cache.stringify(), which defaults to JSON.stringify(). Use a custom parser like flatted to cases like circular object references.
• mSet({ key1: 1, key2: 2, ...}, maxAge): Promise<void>
  • Sets multiple key-value pairs in the cache at one time.
• mSetObjects({ key1: { obj: 1 }, key2: { obj: 2 }, ...}, maxAge): Promise<void>
  • Sets multiple key-value pairs in the cache at one time, where the value is an object. Passes the values through cache.stringify(), see cache.setObject();
• get(key): Promise<string | number | null | undefined>
  • Returns a value for a key.
• getObject(key): Promise<Object | null | undefined>
  • Returns an object value for a key. Passes the values through cache.parse(), which defaults to JSON.parse(). Use a custom parser like flatted to cases like circular object references.
• mGet([key1, key2, ...]): Promise<{key:string | number | null | undefined}?>
  • Returns values for multiple keys, results are in the form of { key1: '1', key2: '2' }.
• mGetObjects([key1, key2, ...]): Promise<{key:Object | null | undefined}?>
  • Returns values as objects for multiple keys, results are in the form of { key1: '1', key2: '2' }. Passes the values through cache.parse(), see cache.getObject().
• peek(key): Promise<string | number | null | undefined>
  • Returns the value for a key without updating its last access time.
• del(key): Promise<void>
  • Removes a value from the cache.
• mDel([key1, key2...]): Promise<void>
  • Removes multiple keys from the cache..
• has(key): Promise<boolean>
  • Returns true if the key exists in the cache.
• incr(key, [amount]): Promise<number>
  • Increments a numeric key value by the amount, which defaults to 1. More atomic in a clustered environment.
• decr(key, [amount]): Promise<number>
  • Decrements a numeric key value by the amount, which defaults to 1. More atomic in a clustered environment.
• reset(): Promise<void>
  • Removes all values from the cache.
• keys(): Promise<Array<string>>
  • Returns an array of all the cache keys.
• values(): Promise<Array<string | number>>
  • Returns an array of all the cache values.
• dump()
  • Returns a serialized array of the cache contents.
• prune(): Promise<void>
  • Manually removes items from the cache rather than on get.
• length(): Promise<number>
  • Return the number of items in the cache.
• itemCount(): Promise<number>
  • Return the number of items in the cache - same as length().
• max([max]): Promise<number | void>
  • Get or update the max value for the cache.
• maxAge([maxAge]): Promise<number | void>
  • Get or update the maxAge value for the cache.
• allowStale([true|false]): Promise<boolean | void>
  • Get or update the allowStale value for the cache (set via stale in options). The stale() method is deprecated.
• execute(command, [arg1, arg2, ...]): Promise<any>
  • Execute arbitrary command (LRUCache function) on the cache, returns whatever value was returned.

example usage

Master

1// require the module in your master thread that creates workers to initialize
2require('lru-cache-for-clusters-as-promised').init();

Worker

1// worker code
2const LRUCache = require('lru-cache-for-clusters-as-promised');
3
4// this is safe on the master and workers. if you need to ensure the underlying
5// LRUCache exists use `await getInstance()` to fetch the promisified cache.
6let cache = new LRUCache({
7  namespace: 'users',
8  max: 50,
9  stale: false,
10  timeout: 100,
11  failsafe: 'resolve',
12});
13
14const user = { name: 'user name' };
15const key = 'userKey';
16
17// using async/await
18(async function() {  
19  // get cache instance asynchronously. this will always be the same underlying cache
20  cache = await LRUCache.getInstance({ /* ...options */ });
21
22  // set a user for a the key
23  await cache.set(key, user);
24  console.log('set the user to the cache');
25
26    // get the same user back out of the cache
27  const cachedUser = await cache.get(key);
28  console.log('got the user from cache', cachedUser);
29
30  // check the number of users in the cache
31  const size = await cache.length();
32  console.log('user cache size/length', size);
33
34  // remove all the items from the cache
35  await cache.reset();
36  console.log('the user cache is empty');
37
38  // return user count, this will return the same value as calling length()
39  const itemCount = await cache.itemCount();
40  console.log('user cache size/itemCount', itemCount);
41}());
42
43// using thenables
44LRUCache.getInstance({ /* ...options */ })
45.then((myCache) => 
46  myCache.set(key, user)
47  .then(() => 
48    myCache.get(key)
49  )
50)

Use a custom object parser for the cache to handle cases like circular object references that JSON.parse() and JSON.stringify() cannot, or use custom revivers, etc.

1const flatted = require('flatted');
2const LRUCache = require('lru-cache-for-clusters-as-promised');
3
4const cache = new LRUCache({
5  namespace: 'circular-objects',
6  max: 50,
7  parse: flatted.parse,
8  stringify: flatted.stringify,
9});
10
11// create a circular reference
12const a = { b: null };
13const b = { a };
14b.a.b = b;
15
16// this will work
17await cache.setObject(1, a);
18
19// this will return an object with the same circular reference via flatted
20const c = await cache.getObject(1);
21if (a == c && a.b === c.b) {
22  console.log('yes they are the same!');
23}

process flow

Clustered cache on master thread for clustered environments

Promisified for non-clustered environments