ClusterAllFailedError: Failed to refresh slots cache.

[achakhalyan]

Hi All. i have following error can You help me ?
[ioredis] Unhandled error event: ClusterAllFailedError: Failed to refresh slots cache.

code is following.
`"use strict";
require('./config/config');
const Redis = require('ioredis');

(async()=> {

const startupNodes = [
    {host: process.env.REDIS_HOST_1, port: process.env.REDIS_PORT },
    {host: process.env.REDIS_HOST_2, port: process.env.REDIS_PORT}
    ];
const options = {
    slotsRefreshTimeout: 2000,
    dnsLookup: (address, callback) => callback(null, address),
    redisOptions: {
        tls: {},
        password: process.env.REDIS_PASS,
    },
};
const cluster = new Redis.Cluster(startupNodes, options);
console.log(cluster);

})();`

[stale]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed after 7 days if no further activity occurs, but feel free to re-open a closed issue if needed.

[iocentos]

Have the same issue. Using an AWS elasticache redis (cluster mode enabled with multiple shards). I get the issue when using in-transit encryption and trying to access behind a NAT instance.

[Dartv]

Same issue here

[peteclark3]

FWIW, I know this issue is re: clustered but I was fighting with this as well (in-transit encryption on, passworded redis) and this connection string finally got me going, I expect it'd work with a cluster as well:

const redis = new Redis({
  host: redisHost,
  port: redisPort,
  password: redisPassword,
  tls: {
    checkServerIdentity: (servername, cert) => {
      // skip certificate hostname validation
      return undefined;
    },
  },
});

[maknahar]

Facing the same issue with 4.11.1 ioredis and redis 3.2.2

[bezzil85]

any updates on this ?

[jsmylnycky]

Same issue here, running latest code:

ioredis @ 4.16.0
elasticache with engine @ 5.0.6

My particular issue was a result of an improperly deployed security group on the ElastiCache cluster, so it was truly unable to connect.

All seems to work ok with the above library versions!

[GSSwain]

The following works for me.

Redis version: 5.0.4 on AWS ElasticCache Clustered with TLS and AUTH enabled.
ioredis version: 4.16.0

Code to connect:

const redis = new Redis.Cluster(
            [{ "host": <ELASTI_CACHE_CONFIGURATION_ENDPOINT> }], {
                dnsLookup: (address, callback) => callback(null, address),
                redisOptions: {
                    tls: true,
                    password: <ELASTI_CACHE_REDIS_AUTH>
                }
            });

Make sure you are in the right VPC, Subnet and Security group to connect to ElastiCache along with the correct IAM role

[ion-willo]

We're having a similar issue.
Redis version: 5.0.4 on AWS ElasticCache Clustered (3 nodes, 1 shard)
ioredis version: 4.16.0

Code to connect:

const Redis = require("ioredis");
var cluster = new Redis.Cluster([
  {
    host: <PRIMARY_RW_ENDPOINT>, port: 6379
  },
  {
   host: <SECONDARY_READERS_ENDPOINT>, port: 6379,
  }
], {
  dnsLookup: (address, callback) => callback(null, address),

  scaleReads: 'slave'
})

This code works, but bypasses the cluster for reads, if I understand it correctly:

const Redis = require("ioredis");
var cluster = new Redis({ host: <PRIMARY_RW_ENDPOINT>, port: 6379 })

I can also connect via redis-cli -h <PRIMARY_RW_ENDPOINT>, so I'm sure VPC, Subnet and Security groups are all correct.

[joebowbeer]

@ion-willo Is your ElastiCache cluster-mode-enabled? If so, I suggest you connect to the single configuration endpoint.

[keyvhinng]

Hi, I have the same issue using AWS ElastiCache & ioredis:

[ElastiCache]
Configuration Endpoint: clustercfg.test1.xxxxxx.use2.cache.amazonaws.com:6379
Engine Version Compatibility: 5.0.6
Parameter Group: default.redis5.0.cluster.on (in-sync)
Engine:Engine: Clustered Redis
Encryption in-transit: Yes
Encryption at-rest: No

[ioredis]
version: 4.17.3

new Cluster([
  {"port":6379,"host":"clustercfg.test1.xxxxxx.use2.cache.amazonaws.com"}
], {
    dnsLookup: (address, callback) => {
            console.log('====[dnsLookup]====');
            return callback(null, address);
    },
    redisOptions: {
            tls: {},
    },
})

Error:
ClusterAllFailedError: Failed to refresh slots cache.

What am I missing ?

[joebowbeer]

I guess you're missing tls.checkServerIdentity and a password, if AUTH is enabled.

I can't vouch for all of the following, but FYI:

return new IORedis.Cluster(endpoints, {
  // https://github.com/luin/ioredis#special-note-aws-elasticache-clusters-with-tls
  // eslint-disable-next-line @typescript-eslint/ban-ts-ignore
  // @ts-ignore(TS2554) https://github.com/luin/ioredis/pull/1143
  dnsLookup: (address, callback) => callback(null, address),
  redisOptions: {
    connectionName,
    password,
    // https://github.com/luin/ioredis#reconnect-on-error
    reconnectOnError: function(err) {
      // Only reconnect when the error contains "READONLY"
      // `return 2` to resend failed command after reconnecting
      return err.message.includes('READONLY')
    },
    showFriendlyErrorStack: true,
    tls: {
      checkServerIdentity: (/*host, cert*/) => {
        // skip certificate hostname validation
        return undefined
      }
    }
  },
  slotsRefreshTimeout: 3000
})

[keyvhinng]

AUTH is not enabled. My current options (and it's still failing):

{
            slotsRefreshTimeout: 3000,
            dnsLookup: (address, callback) => callback(null, address),
            redisOptions: {
                    showFriendlyErrorStack: true,
                    tls: {
                          checkServerIdentity: (/*host, cert*/) => {
                            // skip certificate hostname validation
                            return undefined
                          }
                    },
            },
}

[joebowbeer]

I suggest setting DEBUG="ioredis:*" and see if anything turns up.
I wonder if it's still trying to do AUTH.

[keyvhinng]

I found my mistake, I'm new into ElastiCache and I didn't realize I need to connect from an EC2 or AWS VPN. I fixed that and it worked.

[hejkerooo]

It's not working with https://github.com/bitnami/bitnami-docker-redis-cluster

fixes like checkServerIdentity returning undefined, DNS lookup, and other proposed fixes.
ioRedis version: 4.28.1

[Davidiusdadi]

The only fix that seems to work for me is to up the slotsRefreshTimeout... :/

I wander why the default timeout value is so "dangerously" low. The response to the slots commands just seems to sometimes take 2 or 3 seconds for me. Too bad then that by the time the slots "response" comes the client has already given up...

[lueenavarro]

AUTH is not enabled. My current options (and it's still failing):

{
            slotsRefreshTimeout: 3000,
            dnsLookup: (address, callback) => callback(null, address),
            redisOptions: {
                    showFriendlyErrorStack: true,
                    tls: {
                          checkServerIdentity: (/*host, cert*/) => {
                            // skip certificate hostname validation
                            return undefined
                          }
                    },
            },
}

This will work without failing if you remove the dnsLookup option.

[kidus60]

set DEBUG="ioredis:*" and was able to see connection was timing out before establishing a secure connection via tls...fixed it by setting slotsRefreshTimeout: 10000.

here's our team's full ioredis config

const client =  new Redis.Cluster(
					[
						{
							port: nconf.get('REDIS:PORT') as any,
							host: nconf.get('REDIS:HOST') as string
						}
					],
					{
						enableAutoPipelining: true,
						slotsRefreshTimeout: 10000,
						dnsLookup: (address, callback) => callback(null, address),
						redisOptions: {
							tls: nconf.get('REDIS:TLS') as any
						}
					}
			  );

	client.on('connect', () => {
		logger.info('connected to redis server');
	});

	client.on('close', () => {
		logger.info('redis connection closed attempting reconnect');
		client.connect();
	});

	client.on('error', (error) => {
		logger.error(`redis connection error ${error}`);
	});

[NikithShetty]

Faced similar issue while running redis cluster inside docker container https://github.com/bitnami/containers/tree/main/bitnami/redis-cluster

Fixed it using natMap option. Refer to solution posted on stackoverflow

[sahilkul95]

@kidus60 your solution worked for me. BTW what does enableAutoPipelining: true actually does?

[devSajan]

i am facing the same issue when using clustered config, i have a none clustered Redis instance so when i am using the normal connection it is works on ec2 vm which is in the same vpc but when deployed on eks with docker i am getting some certificate file not found error.

here is my connection config

const options = {
  port: config.redis.port, // Redis port
  host: config.redis.host, // Redis host
  family: 4, // 4 (IPv4) or 6 (IPv6)
  password: config.redis.password,
  db: MAIN_REDIS_DB,
  enableAutoPipelining: true,
  enableOfflineQueue: false,
  tls :{}
}
export const client = new Redis(options);

ioredis version = ^4.28.2
redis engine = 7.0.7

[KaranMane027]

I am also facing the same issue below is my configuration setup
"redis": "^4.5.1"
"ioredis": "^5.2.4"

try { return new Redis.Cluster( [ { host:${redisUrl}, port: process.env.REDIS_PORT, }, ], { dnsLookup: (address, callback) => callback(null, address), redisOptions: {}, clusterRetryStrategy: (times) => { if (times > 1) { console.log("Cluster Memory DB redis connection error"); return new Error("Cluster Memory DB redis connection error"); } else return 10; }, } ); }