Getting EMFILE error on @aws-sdk/client-s3

[sumangalam17]

Checkboxes for prior research

• I've gone through Developer Guide and API reference
• I've checked AWS Forums and StackOverflow.
• I've searched for previous similar issues and didn't find any solution.

Describe the bug

trying to call this function from sendMessage and getting error like below

const saveToS3 = async (bucketName,data) => {

  const uploadParams = {
    Bucket: bucketName,
    Body: JSON.stringify(data)
  };

  const s3 = new S3({ maxAttempts: 3 });

  await s3.putObject(uploadParams)
    .then((data) => {
      logger.info(
        `${componentName}: saveToS3: Successfully uploaded file: `, data);
      return data;
    })
    .catch(err => {
      logger.error(
        `${componentName}: saveToS3: There was an error uploading file: `, err);
    });
}

error :

{
    "errno": -24,
    "code": "EMFILE",
    "syscall": "getaddrinfo",
    "hostname": "s3.us-east-1.amazonaws.com",
    "$metadata": {
        "attempts": 1,
        "totalRetryDelay": 0
    },
    "stack": "Error: getaddrinfo EMFILE us-east-1.amazonaws.com\n    at GetAddrInfoReqWrap.onlookup [as oncomplete] (node:dns:108:26)"
}

tried every possibilities mention in this issue #4345
like calling client globally
using destroy method
setting connection 1 and all

SDK version number

@aws-sdk/package-name@version, 3.418.0

Which JavaScript Runtime is this issue in?

Node.js

Details of the browser/Node.js/ReactNative version

types/node -v 20.2.1

Reproduction Steps

trying to call this function many time from send message

Observed Behavior

n/A

Expected Behavior

n/A

Possible Solution

No response

Additional Information/Context

No response

[ajredniwja]

Hi @sumangalam17 thanks for opening this issue, seems like you are using v2 of the SDK and not using ``@aws-sdk/client-s3.

I would recommend just using try/catch with await.

Can you try to change your code to something like below

for V2

const AWS = require('aws-sdk');

const s3 = new AWS.S3({ maxAttempts: 3 });

const saveToS3 = async(bucketName, data) => {
    const uploadParams = {
        Bucket: bucketName,
        Body: JSON.stringify(data),
        Key: "file.json"
    };

    try {
        const response = await s3.putObject(uploadParams).promise();
        console.log(`Successfully uploaded the file: `, response);
        return response;
    } catch (err) {
        console.error(`Error: uploading`, err);
        throw err;
    }
}

V3 would look like:

const { S3Client, PutObjectCommand } = require("@aws-sdk/client-s3");

const s3 = new S3Client({ region: "YOUR_REGION" });

const saveToS3 = async (bucketName, data) => {
    const uploadParams = {
        Bucket: bucketName,
        Body: JSON.stringify(data),
        Key: "file.json"
    };

    try {
        const response = await s3.send(new PutObjectCommand(uploadParams));
        console.log(`Successfully uploaded the file: `, response);
        return response;
    } catch (err) {
        console.error(`Error: uploading`, err);
        throw err;
    }
}

I tried to call this function several times and wasnt able to reproduce the error.

(async() => {
    try {
        const bucketName = "bucket-name";

        for (let i = 0; i < 1000; i++) {
            const data = { key: `xyz-${i}` }; 
            await saveToS3(bucketName, data);
        }

    } catch (error) {
        console.error("Error uploading:", error);
    }
})();

Please let us know if that doesnt help. thanks

[sumangalam17]

its not due to

I tried to call this function several times and wasnt able to reproduce the error.

(async() => {
    try {
        const bucketName = "bucket-name";

        for (let i = 0; i < 1000; i++) {
            const data = { key: `xyz-${i}` }; 
            await saveToS3(bucketName, data);
        }

    } catch (error) {
        console.error("Error uploading:", error);
    }
})();

Please let us know if that doesnt help. thanks

actually its not due to calling that function many time its due to we are triggering sendMessage and that triggering this function and calling client and function both.

I tried using V3 and try and catch also but still its not working.

[jameshartt]

Definitely seen this when we recently upgraded our v3 AWS dependencies recently. Had to roll back that upgrade. We seemed to see this consistently on many services. Haven't had the time to properly investigate it. But it looked like potentially TCP connections exceeding whatever a given Lambda can cope with. Only happened on Lambda functions that were being called a bunch.

[DesAWSume]

I am on the same boat here, when using this Javascript SDKv3 s3 client. running in Lambda
It randomly throw this error

{
    "errorType": "Error",
    "errorMessage": "bind EMFILE 0.0.0.0",
    "code": "EMFILE",
    "errno": -24,
    "syscall": "bind",
    "address": "0.0.0.0",
    "stack": [
        "Error: bind EMFILE 0.0.0.0",
        "    at __node_internal_captureLargerStackTrace (node:internal/errors:496:5)",
        "    at __node_internal_exceptionWithHostPort (node:internal/errors:671:12)",
        "    at node:dgram:364:20",
        "    at process.processTicksAndRejections (node:internal/process/task_queues:83:21)"
    ]
}

And going through my code, always using await with the PutObjectToS3 function. Can some investigate this?

[praneetrattan]

We have been facing this issue too, but with DynamoDb. We had a lambda being called multiple times concurrently (~503) and this lambda calls DynamoDB for a query. Out of those 503 lambdas, about 93 failed with the error:

{
  "errorType": "Error",
  "errorMessage": "getaddrinfo EMFILE dynamodb.us-east-1.amazonaws.com",
  "code": "EMFILE",
  "errno": -24,
  "syscall": "getaddrinfo",
  "hostname": "dynamodb.us-east-1.amazonaws.com",
  "$metadata": {
    "attempts": 1,
    "totalRetryDelay": 0
  },
  "stack": [
    "Error: getaddrinfo EMFILE dynamodb.us-east-1.amazonaws.com",
    "    at GetAddrInfoReqWrap.onlookup [as oncomplete] (node:dns:108:26)"
  ]
}

Any known solution for this? Some forums mentioned it might be a problem with how connections are handled in node. We just upgraded to V3 and as per docs AWS_NODEJS_CONNECTION_REUSE_ENABLED is set to true by default.
https://docs.aws.amazon.com/sdk-for-javascript/v3/developer-guide/node-reusing-connections.html

[Altane-be]

I'm also having this issues since passing to v3 for a number of lambda when initializing a new instance of DynamoDb class. Could we have an update on this please ?

[mahdiridho]

I get lot of emfile messages when try to call any api gw endpoint from lambda. Still stuck till now. Most of the articles I have tried, but no luck

[randomhash]

Describe the bug

Error: getaddrinfo EMFILE s3.eu-central-1.amazonaws.com
    at GetAddrInfoReqWrap.onlookup [as oncomplete] (node:dns:107:26)
    at GetAddrInfoReqWrap.callbackTrampoline (node:internal/async_hooks:128:17)
Processing large amounts of data leads to such errors with s3/dynamodb/possible other clients

SDK version number
3.362.0

Which JavaScript Runtime is this issue in?
Node.js

Details of the browser/Node.js/ReactNative version
lambda node 18

Reproduction Steps
That is tough one. Large amount of data processed within step functions

Dynamodb client 1:

export function getClient(credentials: Creds, forceRecreation: boolean = false): DynamoDBDocument {
  if (forceRecreation || !documentClient || areCredentialsExpired()) {
    documentClient = DynamoDBDocument.from(
      new DynamoDBClient({
        requestHandler: new NodeHttpHandler({
          httpsAgent: new Agent({
            keepAlive: true,
            maxSockets: Infinity,
          }),
        }),
        credentials: {
          accessKeyId: credentials.accessKeyId,
          secretAccessKey: credentials.secretAccessKey,
          sessionToken: credentials.sessionToken,
          expiration: credentials.expiration as Date,
        },
      })
    );

    cachedCredentials = credentials;
  }

  return documentClient;
}

Dynamodb client 2:

export const documentClient = DynamoDBDocument.from(
  new DynamoDBClient({
    logger: console,
    retryStrategy,
  })
);

s3 clients:

import {retryStrategy} from '../retry/index.js';

const s3download = new S3({
  region: process.env.AWS_REGION,
  logger: console,
  forcePathStyle: true,
  requestHandler: new NodeHttpHandler({
    httpsAgent: new Agent({
      timeout: 5000,
      maxSockets: 1000, // default 50
      // keepAlive is a default from AWS SDK. We want to preserve this for
      // performance reasons.
      keepAlive: true,
      keepAliveMsecs: 1000, // default unset,
    }),
    connectionTimeout: 1000,
    requestTimeout: 5000, // default 0
  }),
  retryStrategy,
});

const s3upload = new S3({
  region: process.env.AWS_REGION,
  logger: console,
  forcePathStyle: true,
  requestHandler: new NodeHttpHandler({
    httpsAgent: new Agent({
      maxSockets: 1000, // default 50
      keepAlive: false,
      keepAliveMsecs: 1000, // default unset,
    }),
    connectionTimeout: 1000,
    requestTimeout: 5000, // default 0
  }),
  retryStrategy,
});

export function getClient(params?: S3ClientConfig) {
  if (params) {
    return new S3(params);
  }

  return s3download;
}

Large amounts of data processed in this way leads to repetetive errors with emfile
s3 clients initiated once, ddb2 instance only 1, 1st instance is being reset on credentials expire

[RanVaknin]

Hi everyone on the thread,

Sorry for the late response.

A lot of folks have commented here and shared info, but the part I'm suspicious about is the actual Lambda handler function itself.

In order to limit the amount of open file descriptors, you can take advantage of lambda's re-use of the execution environment.

• Lambda Runtime Environment
• Container Reuse in lambda

By instantiating the SDK client outside of the handler's codepath, you will ensure you have one sdk client and that connections will be manged more efficiently.

I added a code example to our docs https://github.com/aws/aws-sdk-js-v3?tab=readme-ov-file#best-practices

Let me know if this helps.
Thanks,
Ran~

Related:
#4345
#5010
#4964

[randomhash]

@RanVaknin hey, just a quick note, any idea on how to organize cross-account communication? Or maybe suitable sample, whenever we have in equation at least 2 aws accounts that share services?
As there are STS credentials with expiration 1 instance per lambda may be a bit problematic, especially whenever we are utilizing long-running tasks with complex data transfers
While making the actual destroy of the client is an option, still, looks like it does not help us to solve the EMFILE while there are lots of action happening within lambda itself

[RanVaknin]

Hi @randomhash ,

Like I mentioned in my comment above, it's hard to advise on a specific scenario without seeing your actual handler code.
Calling .destroy() is indeed a workaround but like you mentioned might not cover all cases. The correct approach is to instantiate the SDK client outside of the handler's codepath, in the function scope so that when the container is reused, there wont be another SDK created in memory.

If you can share a code example of what your application does we might be able to help you re-arrange your code in a way that avoids recreation of the SDK clients.

Thanks,
Ran~

[github-actions]

This issue has not received a response in 1 week. If you still think there is a problem, please leave a comment to avoid the issue from automatically closing.

[github-actions]

This thread has been automatically locked since there has not been any recent activity after it was closed. Please open a new issue for related bugs and link to relevant comments in this thread.