Hi,

First all! Thanks a lot for the lib. I have been looking for one and I hope the official strapi team starts looking at this 👍 .

I am trying to get it working on my project. I have set up the middleware and add the model.

module.exports = ({ env }) => ({
  settings: {
    cache: {
      enabled: true,
      models: ['ratings']
    },
  },
});

I don't see the message.

Anything that I am doing wrong? Following is the debug info:

Please let me know if you need further info and grateful if you can help :)

Cheers,
Jeremy

When editing a collection type in the admin cache busting isn't working anymore because the pattern variable in the bust function isn't using pluralize since v1.0.9, which makes it so that the pattern isn't matching the cache keys anymore. Everything works fine for single types though, since the pluralize function is avoided in this context (as it should).

v1.0.8:

const pattern     = new RegExp(`^/${pluralize(model)}`);

v1.0.9

const pattern     = new RegExp(`^/${model}`);

Using a bit of your code from function bustAdmin, I'm able for figure out if the pattern should be pluralized or not (This works for both single types and collection types):

const bust = (model) => async (ctx, next) => {
  const modelCfg = _.find(toCache, (cfg) => {
    return cfg.model === (cfg.singleType ? model : pluralize(model));
  });

  const pattern     = new RegExp(`^/${modelCfg.model}`);
  const keys        = await cache.keys() || [];

  await next();

  if (!_.inRange(ctx.status, 200, 300)) return;

  await Promise.all(
    _.filter(keys, k => pattern.test(k))
      .map(k => cache.del(k))
  );
}

Hi, is it possible to bust the cache for a specific model by using id?

I need some instructions on how I can do it inside of my controller.

Thanks!

Description: Strapi-middleware-cache is not updating/busting the Redis cache accordingly on Strapi v3.0.6.

Issue: We are using Strapi-middleware-cache plugin and updated our middleware.js accordingly.
We are using Redis storage for caching collections. We want to bust/flush & update the cache accordingly for each update/create/delete any collection's entries.

But we found the Redis cache is getting bust/flush only with Redis server restart (each time we had to manually restart Redis).

We are following this documentation: https://strapi.io/blog/caching-in-strapi-strapi-middleware-cache

Please see our middleware.js code below:

module.exports = ({ env }) => ({
  settings: {
    cache: {
      enabled: true,
      type: "redis",
      maxAge: 3600000,
      logs: true,
      populateContext: true,
      populateStrapiMiddleware: true,
      enableEtagSupport: true,
      models: ["cars", "tests"],
      redisConfig: {
        host: "127.0.0.1",
        port: 6379,
      },
    },
  },
});

And we updated the cars controller accordingly on this file: api/cars/controllers/cars.js

module.exports = {
  async index(ctx) {
    console.log("Inside cars controller--------");
    ctx.middleware.cache.store;
    await ctx.middleware.cache.bust({ model: "cars" });
  },
};

Actual Issue: Cache is not getting flush/bust automatically on create/update/delete operations. We found the Redis cache is getting bust/flush only with Redis server restart, each time when any entry is update/deleted/created, we had to manually restart Redis.

Expected: We want to bust/flush & update the cache accordingly for each update/create/delete operations for any collection's entries. We don't need to restart Redis server to flush/update the cache.

System:
OS: MAC OS Catalina 10.15.7
Strapi: Strapi v3.0.6
strapi-middleware-cache: "^1.5.0"
redis_version: v6.2.5

Hi!

I was wondering if you would consider adding an option to bypass serving the cache entry when matched in the cache store. I'm working on implementing front-end cache using the ETag and If-None-Match headers. I've already added that layer to Strapi and in addition I'd like to leverage the cache store from this middleware by using the cache entry to create the ETag instead of querying the database every time.

I have Strapi setup and in my frontend all the interaction with Strapi happens via GraphQL. When using this middleware will caching work?

Reproduction Steps:

1. Create a new item e.g. a new page under pages
2. Get the list of pages e.g. /pages, the new unpublished item is missing as expected
3. Publish the new page
4. Get the list of pages /pages, the newly published item is missing

The same is true when changing state from published to unpublished. The cache is not evicted and as a result the item is still visible in the list eventhough it is not published

The workaround for this issue is to publish and then modify in order to evict the cache

The config I've setup in config/middleware.js does not match the config returned by getCacheConfig

I have this model setup in my config:

{
  model: "contracts",
  hitpass: false,
  injectDefaultRoutes: false,
  routes: [
    {
      path: "/contracts/:contractType/:countryLocale",
      method: "GET",
      paramNames: ["contractType", "countryLocale"],
    },
  ],
}

Calling getCacheConfig("contracts") returns the following, notice how the paramNames does not match

{
  singleType: false,
  hitpass: false,
  injectDefaultRoutes: false,
  headers: [],
  maxAge: 3600000,
  model: 'contracts',
  routes: [
    {
      path: '/contracts/:contractType/:countryLocale',
      method: 'GET',
      paramNames: [ 'contractType' ]
    }
}

Are there any plans for v4 since Strapi v4 is in beta currently?

I have a custom API which get's the data from different content types. The inbuilt mechanism doesn't automatically refreshes the cache on custom routes if the content type data is changed. As per documents ctx is required to bust the cache.

await ctx.middleware.cache.bust({ model: 'posts' });

But ctx is not available in models. How can I manually bust the cache from models life cycle hooks like beforeDelete

Hey!

Firstly, many thanks for this very useful package! 👍

I've added the following code in at config/middleware.js

module.exports = ({ env }) => ({
    settings: {
      cache: {
        enabled: true,
        models: ['grid'], 
      }
    }
  })

I can see the mounting and the using storage engine logs but I have no log from the caching routes.
I also tried with grids, nothing happens.

Here are the logs:

$ strapi develop
[2020-11-23T20:27:40.633Z] debug [Cache] Mounting LRU cache middleware
[2020-11-23T20:27:40.634Z] debug [Cache] Storage engine: mem

Missing this log :
debug [Cache] Caching route /grids/* [maxAge=3600000]

Any idea?

Many thanks :)

I have a MongoDB server holding data which I don't change much. Most queries are to read. Is it possible to set maxAge for lru cache forever, thus the lru cache will never expire by itself?

strapi: 3.6.6
strapi-middleware-cache: 2.1.0

i have city and building models, i want to clear building cache when i update or delete city model.
in building/models/building/settings.json, i have set {attributes: {city: {model: "city"}}}
in middleware.js, i have set {clearRelatedCache: true}, it's not clear building cache when i edit or delete city model.

document said (this clears the whole cache for the related models), is there any way to set related models?

Environment:

• node: v16.13.0
• npm: 7.17.0
• storage engine: redis

Hi, I'am having a request from the front with _where query param but in the cache its stored like object and all the _where requests will be the same.

When I have investigated the redis instance found in key _where[object Object]
I guess you are missing toString() to evaluate the _where object otherwise all the requests that uses _where will be the same for different objects.

Hello 👋

We are using the library and so far is working great, but we came up with a possible scenario: If the connection with Redis is lost for whatever reason (maybe the cache goes down, or the connection pipe gets broken), then the library will try to fetch data from Redis and will get stuck forever (which is far from ideal). It would be great if the library could handle this case.

We think that this could be avoided by doing two things:

1. Change the status flag to use the status property of the ioredis' client.
2. Including a cache timeout, just in case.

We have some work in progress in a fork, so I will open a draft PR, but I would like to hear your thoughts on this one.

Keep up the good work with the library! :D

this way i setup the cache

this is the route in routes.json

on running strapi its caching route only default one

please guide

I'm using Strapi with the GraphQL plugin. Is there any chance it will get supported? GraphQL is used with POST.

The cache doesn't return the updated data.
I used the default memory cache and following is my cache config in/config/middleware.js
cache: {
enabled: true,
enableEtagSupport: true,
enableXCacheHeaders: true,
withKoaContext: true,
withStrapiMiddleware: true,
clearRelatedCache: true,
maxAge: 3.154e10,
models: [
{
model: 'sqore-posts',
hitpass: false,
singleType: true,
routes: ['/sqore-posts/:slug', '/sqore-posts'],
},
],
},

After updating the content, updated data is not from API and only old cache data when calling /sqore-posts/:slug.

So I tried to clear or update the cache of model "sqore-posts" in the sqore-posts model.
but I don't know how to clear or update the cache.
I used this code in /models/sqore-posts.js.

afterUpdate: async (entry) => {
    console.log('---------------sqore-posts afterUpdate')
    const cacheConf = await strapi.middleware.cache.getCacheConfig('sqore-posts')
    await strapi.middleware.cache.clearCache(cacheConf)
},

I am looking for your help.
Thanks.

I am running half-a-dozen Strapi apps off a DigitalOcean droplet with Redis locally installed for caching purposes. I noticed today that two of those sites were showing each other's Strapi data in their front-end client applications. I quickly narrowed this down to an issue with key collisions in Redis. I was able to resolve the issue by using the redisConfig option to specify a db for each Strapi app. But I thought I would call it to your attention anyway in case this was an issue you were not aware of and might have an idea of how to resolve without needing to manually specify a different Redis database for each application. Thanks!

Due the re-write of of the middleware to version 2, the old tests had to be disabled

The latest working versions of the tests is here

Here's an overview

 Caching
    Collection types
      ✓ should let uncached requests go through
      ✓ following request should be cached
      ✓ following request should have response status 304
      ✓ caches based on headers
      ✓ doesnt cache models not present in the config
      ✓ busts the cache on a POST resquest
      ✓ busts the cache of a related model (via direct relation) on a POST resquest
      ✓ busts the cache of a related model (via relation in component) on a POST resquest
      ✓ busts the cache on an admin panel POST resquest
      ✓ busts the cache on an admin panel POST publishing change
      ✓ busts the cache on a PUT resquest
      ✓ busts the cache of a related model (via direct relation) on a PUT resquest
      ✓ busts the cache of a related model (via relation in component) on a PUT resquest
      ✓ busts the cache on an admin panel PUT resquest
      ✓ busts the cache on an admin panel PUT publishing change
      ✓ busts the cache on a DELETE resquest
      ✓ busts the cache of a related model (via direct relation) on a DELETE resquest
      ✓ busts the cache of a related model (via relation in component) on a DELETE resquest
      ✓ busts the cache on an admin panel DELETE resquest
      ✓ busts the cache on an admin panel DELETE publishing change
      when an ID is specified on a POST request
        ✓ doesn't bust the cache for other IDs
      when an ID is specified on a PUT request
        ✓ doesn't bust the cache for other IDs
      when an ID is specified on a DELETE request
        ✓ doesn't bust the cache for other IDs
    Single types
      ✓ caches non pluralized endpoints
      ✓ doesnt cache pluralized endpoints
      ✓ busts the cache on a POST resquest
      ✓ busts the cache on an admin panel POST resquest
      ✓ busts the cache on a PUT resquest
      ✓ busts the cache on an admin panel PUT resquest
      ✓ busts the cache on a DELETE resquest
      ✓ busts the cache on an admin panel DELETE resquest

Ideally the tests would be an end-to-end setup which spins up an actual Strapi instance to test against.

cc @stafyniaksacha

Hi,

I integrated the strapi middleware cache to my project and added the model like categories. Below is my cache settings in middleware.js under config
cache: {
enabled: true,
type: 'redis',
maxAge: 3600000,
models: [
{
model: 'Categories',
maxAge: 1000000,
}
],
populateStrapiMiddleware: true,
logs:true
}
First time when GET /Categories api is invoked it asked for bearer token. Next time when it is cached authentication is totally ignored. If I remove auth header no forbidden message is shown instead data is returned giving access to anybody with api. Am I missing any steps? How to overcome this problem so that cache middleware is not invoked before auth is verified.

The cache is not cleared when we change data through the admin panel

So in looking at your code, it would appear you are specifically pulling out config options that may not always apply.

Looking through the ioredis config options (and what I am currently testing) is what is known as the Redis HA config or Sentinel system: https://github.com/luin/ioredis#sentinel

While it's a tad difficult to see in this example screenshot:

Basically there are 3 servers, running 6 services:

Server 1: Redis Server + Sentinel (Master)
Server 2: Redis Server + Sentinel (Slave)
Server 3: Redis Server + Sentinel (Slave)

When configuring the ioredis for the cache, we would simply pass the Sentinel addresses like:

var redis = new Redis({
  sentinels: [
    { host: "192.168.10.41", port: 26379 },
    { host: "192.168.10.42", port: 26379 },
    { host: "192.168.10.43", port: 26379 },
  ],
  name: "redis-primary",
});

And thus there would be no root host, port, pass, ect and would be contained within the sentinels array. Likewise for the name as sentinels can manage multiple Redis "clusters" (tad confusing here I know)

I will play around locally and see if I can find a solution to submit a PR unless you already have some ideas. Personally I think just simply having a config option for redisConfig: {} to allow someone to use whatever they want would be better then having the config options in the root of the settings object.

FYI I am also testing on the latest Strapi next tagged release which is what will be 3.0.0-rc.1 when they release it (Release Candidate for the Stable)

Hi, it seems like strapi.middleware.cache is not being initialized as specified in the readme. This is my config:

cache: {
      enabled: true,
      maxAge: 3600000,
      populateStrapiMiddleware: true,
      models: ["contract"]
}

but trying to call functions in lifecycle through strapi.middleware.cache just returns xx is not a function

Here is the model-file in question:

module.exports = {
  lifecycles: {
    afterUpdate: async (result) => {
      const { cache } = strapi.middleware
      console.log("Cache: ", cache);
      console.log(cache.getCacheConfig("contract"));

Am I missing something?

Description: Failing Redis connectivity when using DEV Redis

Steps to reproduce:

1. Install this Library on the Strapi app
2. To connect this library to Redis - Use Per-Model configuration mentioned on strapi-middleware-cache NPM docs
3. Try to connect Strapi app to any DEV instance of Redis server (stop your local Redis server on your machine)
4. Run the app
5. Throws below error:

error [Cache] Redis connection failed
[2021-05-14T16:38:20.740Z] error Error: connect ECONNREFUSED 127.0.0.1:6379

So though my Strapi app is connected to DEV Redis but this middleware is still trying to connect to local Redis (127.0.0.1:6379) by default and since my local Redis is stopped, it throws above connectivity error.

We are using below Per-Model Configuration mentioned on https://www.npmjs.com/package/strapi-middleware-cache

module.exports = ({ env }) => ({
  settings: {
    cache: {
      enabled: true,
      type: 'redis',
      maxAge: 3600000,
      models: [
          'cars',
          'articles',
          'test',
          'trucks'
        ]
    }
  }
});

Notes: This middleware is working fine when Strapi app is connected to local Redis server. Issue is only when Strapi app is connected to DEV Redis

Expected Behavior: This middleware should work fine though we use Local/DEV/UAT/PROD Redis server's
Please let us know what configuration we need to pass to this middleware to use the correct Redis what Strapi is connected to?

Hi!

I'm having issues with cache for related models not being busted.
Workflow is as follows:

1. Query /regions/germany, /regions/germany?& is added to cache
2. Change other model which has one-to-one relation with regions through strapi-admin, which should bust the cache
3. Relations cache is busted with wildcard as we don't know params for related entries.

Issue is that in https://github.com/patrixr/strapi-middleware-cache/blob/master/lib/utils/getRouteRegExp.js#L27

  // wildcard: clear all routes
  if (wildcard) {
    let pattern = route.path;
    for (const paramName of route.paramNames) {
      pattern = pattern.replace(`:${paramName}`, '([^/]+)');
    }

    return [new RegExp(`^${pattern}\\?`)];
  }

pattern.replace(':${paramName}', '([^/]+)'); should be changed to pattern.replace(paramName, '([^/]+)'); as paramName already contains :.

Or we can map through paramNames and remove : and later add it when replacing.

I prepared PR with the needed fix #77 @patrixr

And many thanks for the great lib, it helped us to decrease loading times significantly!

The matching of params when creating the regex is breaking since params extracted from the route will have format [":param1", ":param2"] while the ones added wont have the : since they are likely used as:

getCacheConfRegExp(cacheConf, {
  param1,
  param2,
});

Even if you explicitly name the params

getCacheConfRegExp(cacheConf, {
  ":param1": param1,
  ":param2": param2,
});

the pattern-replace will fail in the loop a few lines down since it then tries to match ::param1 which doesn't exist in the route

pattern = pattern.replace(`:${paramName}`, params[paramName]);

Hello, thank you kindly for putting together this package. I'm very new to redis and I'm having a hard time connect to the redis instance I set up on Redislabs

My /config/middleware.js files looks like this (some values have been changed):

  settings: {
    cache: {
      enabled: true,
      type: 'redis',
      maxAge: 3600000,
      models: ['posts'],
      redisConfig: {
        url: 'redis://redis-5555.c8.us-east-1-4.ec2.cloud.redislabs.com:5555',
        name: 'content',
        password: 'secret_password',
      },
    },
  },
});

But I keep getting the following error:

Error: connect ECONNREFUSED 127.0.0.1:6379 at TCPConnectWrap.afterConnect [as oncomplete]

Hi, as above, :-)

Thanks

How i can disable pluralize plugin or cache single types like menu, main-page, news-page? When i add menu to models array plugin modify this to menus

Example:

[2022-04-03T11:17:27.434Z] info [cache] Mounting LRU cache middleware
[2022-04-03T11:17:27.435Z] info [cache] Storage engine: mem
[2022-04-03T11:17:27.441Z] debug [cache] Register dealers routes middlewares
[2022-04-03T11:17:27.442Z] warn [cache] POST /dealers has no matching endpoint, skipping
[2022-04-03T11:17:27.443Z] warn [cache] DELETE /dealers/:id has no matching endpoint, skipping
[2022-04-03T11:17:27.443Z] debug [cache] PUT /dealers/:id purge
[2022-04-03T11:17:27.445Z] debug [cache] GET /dealers recv maxAge=3600000
[2022-04-03T11:17:27.445Z] debug [cache] GET /dealers/:id recv maxAge=3600000
[2022-04-03T11:17:27.474Z] debug [cache] Register admin routes middlewares
[2022-04-03T11:17:27.474Z] debug [cache] POST /content-manager/single-types/:model/actions/publish purge-admin
[2022-04-03T11:17:27.474Z] debug [cache] POST /content-manager/single-types/:model/actions/unpublish purge-admin
[2022-04-03T11:17:27.474Z] debug [cache] POST /content-manager/collection-types/:model purge-admin
[2022-04-03T11:17:27.475Z] debug [cache] POST /content-manager/collection-types/:model/:id/actions/publish purge-admin
[2022-04-03T11:17:27.475Z] debug [cache] POST /content-manager/collection-types/:model/:id/actions/unpublish purge-admin
[2022-04-03T11:17:27.475Z] debug [cache] POST /content-manager/collection-types/:model/actions/bulkDelete purge-admin
[2022-04-03T11:17:27.476Z] debug [cache] PUT /content-manager/single-types/:model purge-admin
[2022-04-03T11:17:27.476Z] debug [cache] PUT /content-manager/collection-types/:model/:id purge-admin
[2022-04-03T11:17:27.476Z] debug [cache] DELETE /content-manager/single-types/:model purge-admin
[2022-04-03T11:17:27.476Z] debug [cache] DELETE /content-manager/collection-types/:model/:id purge-admin

[2022-04-03T11:19:13.438Z] debug [cache] GET /dealers MISS
[2022-04-03T11:19:13.441Z] debug GET /dealers (178 ms) 200
[2022-04-03T11:19:20.462Z] debug [cache] GET /dealers HIT
[2022-04-03T11:19:20.465Z] debug GET /dealers (4 ms) 200
[2022-04-03T11:19:23.846Z] debug PUT /dealers/5fbc20181740b2002cf9deb4 (326 ms) 200
[2022-04-03T11:19:25.357Z] debug [cache] GET /dealers MISS
[2022-04-03T11:19:25.360Z] debug GET /dealers (191 ms) 200
[2022-04-03T11:19:28.698Z] debug GET /content-manager/collection-types/application::dealer.dealer/5fb79f7ebcd3da002c163063 (250 ms) 200
[2022-04-03T11:19:32.228Z] debug PUT /content-manager/collection-types/application::dealer.dealer/5fb79f7ebcd3da002c163063 (480 ms) 200
[2022-04-03T11:19:34.019Z] debug [cache] GET /dealers HIT
[2022-04-03T11:19:34.020Z] debug GET /dealers (2 ms) 200

using PUT does purge the cache, but changing an entry in the UI does not :(
Reverting back to 1.5, it does purge from the UI.

Node: 14.18.0
Strapi: 3.6.0

Strapi has some grammatical logic around pluralization for its apis so words like "elsewhere" or "offsite" don't end up having s appended to the end of it. Ideally this app would utilize the same code to pluralize its apis

Hello!

Is strapi cache compatible with strapi 4?

I have a model with 3 fields containing {id(auto-generated), username, email}

I have modified the default GET endpoint to "username" instead of "id" to search specific record based on username like
GET /profiles/someusername

But for PUT/DELETE I'm using "id" as the endpoint PUT/DELETE /profiles/random_generated_id

Whenever I use PUT/DELETE, data changes in the database, the previously cached data of different endpoint is not busting. I'm thinking of extending those PUT/DELETE API and manually deleting the specific endpoint's cache.

Help me get out of this situation.

I'm trying to use it in redis mode but the log keeps saying
warn [cache] the store is not ready
However, I can connect to that redis with redis-cli -a <host> -p <port> -a <password>

here's my config

module.exports = ({env}) => ({
  settings:{
    audit:{
      enabled: true
    },
    cache:{
      enabled: true,
      logs: true,
      maxAge: 3600000,
      clearRelatedCache: true,
      withStrapiMiddleware: true,
      type: "redis",
      models: [...],
      redisConfig:{
        host: host,
        port: port,
        password: password
      }

    }
  }
})

Let consider this relation: A Book can have multiple authors
Steps:

1. My endpoint for fetching the books is HTTP://localhost:1337/books. Data is being cached here. This endpoint also fetches the authors data from the authors collection.
2. Go to the admin panel and update any author from that book.
3. Call the get books endpoint, it brings data from the cache but the updated author info is not available.
4. Go to the admin panel and update any book
5. Now the cache has been busted.

How can this be achieved? Is there a way to manually bust the cache via another endpoint just like graphql has exposed its playground? Help needed.

PS: I am using strapi with mongo.

When clearRelatedCache is enabled and a model contains a field which is a file collection, the following crash occurs:

TypeError: Cannot read property 'kind' of undefined
      at /Users/foo/myproject/node_modules/strapi-middleware-cache/lib/index.js:195:51
      at /Users/foo/myproject/node_modules/lodash/lodash.js:4967:15
      at baseForOwn (/Users/foo/myproject/node_modules/lodash/lodash.js:3032:24)
      at /Users/foo/myproject/node_modules/lodash/lodash.js:4936:18
      at Function.forEach (/Users/foo/myproject/node_modules/lodash/lodash.js:9410:14)
      at clearCache (/Users/foo/myproject/node_modules/strapi-middleware-cache/lib/index.js:194:13)
      at async /Users/foo/myproject/node_modules/strapi-middleware-cache/lib/index.js:248:9
      at async bustAdmin (/Users/foo/myproject/node_modules/strapi-middleware-cache/lib/index.js:318:11)
      at async /Users/foo/myproject/node_modules/strapi/lib/middlewares/responses/index.js:9:9
      at async cors (/Users/foo/myproject/node_modules/@koa/cors/index.js:95:16)
      at async /Users/foo/myproject/node_modules/strapi/lib/middlewares/logger/index.js:56:11
      at async /Users/foo/myproject/node_modules/strapi/lib/middlewares/responseTime/index.js:17:9
      at async /Users/foo/myproject/node_modules/strapi/lib/services/metrics/middleware.js:29:5
      at async /Users/foo/myproject/node_modules/strapi/lib/Strapi.js:325:9

In my project I have a field called imageGallery which looks like this:

    "imageGallery": {
      "collection": "file",
      "via": "related",
      "allowedTypes": [
        "images"
      ],
      "plugin": "upload",
      "required": false,
      "pluginOptions": {}
    }

I think the problem is in the getRelatedModels function, line 102ff:

    if (attr.collection) {
      relatedModels[attr.collection] = models[attr.collection];
    } else if (attr.model && attr.model !== 'file') {
      relatedModels[attr.model] = models[attr.model];
    }

In the first case, when attr.collection is true, I think it should also check that the collection != 'file'

I got the issue when setting enableEtagSupport to be true but no etag included in the response header
What else do I need to do

below are my settings

cache: { enabled: true, type: 'mem', enableEtagSupport: true, enableXCacheHeaders: true, maxAge: CACHE_AGE_MEDIUM, models: [...], ...}

Hi,

Is it possible to override the hostname of the strapi-middleware-cache plugin? I am using Redis container in my docker-compose.yml and by default Strapi calls host 127.0.0.1 instead of my container name. I would like replace 127.0.0.1 by redis

Because I get this error:

error [cache] redis connection failed {"errno":-111,"code":"ECONNREFUSED","syscall":"connect","address":"127.0.0.1","port":6379}

I tried to add REDIS_HOST env variable but same problem

Hi,

With the latest version of Strapi (3.4.1) I noticed the cache isn't busted anymore when saving/updating an entry in the CMS.
I did notice in the source (https://github.com/patrixr/strapi-middleware-cache/blob/master/lib/index.js#L189) that the admin REST endpoint refer to /content-manager/explorer/:scope and that Strapi uses /content-manager/collection-types/:scope.

I'm not fully sure if these routes have changed in the latest version of Strapi, but I think this prevents the busting of the cache as those do not match anymore.

i'm getting 500 errors submitting posts to strapi because this extension seems to be choking on the unique id field

[2021-08-02T03:03:18.800Z] debug GET /content-manager/collection-types/application::posts.posts?page=1&pageSize=20&_sort=date:DESC (97 ms) 200
[2021-08-02T03:03:24.820Z] debug POST /content-manager/uid/generate (88 ms) 200
[2021-08-02T03:03:26.405Z] debug POST /content-manager/uid/generate (48 ms) 200
[2021-08-02T03:03:26.950Z] debug POST /content-manager/uid/check-availability (35 ms) 200

  TypeError: Cannot read property 'kind' of undefined
      at /workspace/node_modules/strapi-middleware-cache/lib/index.js:195:51
      at /workspace/node_modules/lodash/lodash.js:4967:15
      at baseForOwn (/workspace/node_modules/lodash/lodash.js:3032:24)
      at /workspace/node_modules/lodash/lodash.js:4936:18
      at Function.forEach (/workspace/node_modules/lodash/lodash.js:9410:14)
      at clearCache (/workspace/node_modules/strapi-middleware-cache/lib/index.js:194:13)
      at processTicksAndRejections (internal/process/task_queues.js:95:5)
      at async /workspace/node_modules/strapi-middleware-cache/lib/index.js:248:9
      at async bustAdmin (/workspace/node_modules/strapi-middleware-cache/lib/index.js:318:11)
      at async /workspace/node_modules/strapi/lib/middlewares/responses/index.js:9:9
      at async cors (/workspace/node_modules/@koa/cors/index.js:95:16)
      at async /workspace/node_modules/strapi/lib/middlewares/logger/index.js:56:11
      at async /workspace/node_modules/strapi/lib/middlewares/responseTime/index.js:17:9
      at async /workspace/node_modules/strapi/lib/services/metrics/middleware.js:29:5
      at async /workspace/node_modules/strapi/lib/Strapi.js:325:9

Hello o/

First off thank you for creating this middleware! I have been testing it for the new Strapi rc versions (working great by the way) and I was wondering if you would be interested to add support for the redis-clustr as I believe this will be very helpful for larger installations of Strapi in a cluster and thus allowing all of them to share the cache.

Please do let me know what you think!

While doing some testing on my own project, I noticed the initial logging could quickly get out of hand (for reference I have 90-ish models in my project, and am only caching about 20% of them)

I'm thinking it might be better to add some kind of config option to disable the verbosity (or at least decrease it) of the logging.

An external call will get the cached data. But how would a route in my controller fetch the same cached data?

Do I need to use axios to call http://localhost:1337/posts. Would be great to cache at the model layer instead of controller.

Example: strapi.services.posts.find()

I am testing out the beta branch as I have a custom route that is currently being cached forever despite updates.

Use case

My courses/:id/children route computes the children of the current course (id). So it is dependent of any course that are children of this item.

Issue

This custom children route only has GET implemented. So the cache is never cleared despite the underlying children that have been updated.

What I need

I want to bust the cache of the children route upon any of the courses are updated/deleted/posted. In essence I want to just clear all children cache anytime any course is changed. I don't need fine grain control.

Alternatives

Looking at the beta docs, it seems I might be able to call the cache object directly within the model hooks. This seems reasonable. Why is the warning about using this feature there?

https://github.com/patrixr/strapi-middleware-cache/tree/beta#strapi-middleware

Description: Redis connection failing when using a separate redis server

This middleware working fine on local with running local redis server.

But I have a separate redis instance running on DEV/UAT/PROD and i connect to it through a different library (not ioredis)

I have used this below configuration:
// config/middleware.js

module.exports = ({ env }) => ({
  settings: {
    cache: {
      enabled: true,
      type: 'redis',
      maxAge: 3600000,
      models: [
        {
          model: 'listings',
          maxAge: 1000000,
        }
      ]
    }
  }
});

My DEV/UAT/PROD redis has a different type: redis-something.

How can i make this middleware work with a different redis server?

So in my research and initial testing with clusters: #3

There seems to be an issue with redis-lru in which clusters are failing to set cache values due to how that package handles keys and pipelines. I'm currently looking at alternatives but we could try to reach out the package author about it.

I did see the following issue (from 2018) that seems slightly related: facundoolano/redis-lru#11

This plugin is amazing but is constructed around strapi v3. Is there any plan or guide for how to use this for strapi v4?

Hello, thanks for this great package! I noticed that if I have a single type model with hasMany relation, updating data in the related model doesn't bust this single model cache.

That's how my relation looks like.

I wonder, is there any way to make it work? Or it would require changes in the middleware code itself?