Amazon MSK Library for AWS Identity and Access Management

License

This project is licensed under the Apache-2.0 License.

Introduction

The Amazon MSK Library for AWS Identity and Access Management enables developers to use AWS Identity and Access Management (IAM) to connect to their Amazon Managed Streaming for Apache Kafka (Amazon MSK) clusters. It allows JVM based Apache Kafka clients to use AWS IAM for authentication and authorization against Amazon MSK clusters that have AWS IAM enabled as an authentication mechanism.

This library provides a new Simple Authentication and Security Layer (SASL) mechanism called AWS_MSK_IAM. This new SASL mechanism can be used by Kafka clients to authenticate against Amazon MSK clusters using AWS IAM.

• Amazon Managed Streaming for Apache Kafka
• AWS Identity and Access Management
• AWS IAM authentication and authorization for MSK

Building from source

After you've downloaded the code from GitHub, you can build it using Gradle. Use this command:

gradle clean build

The generated jar files can be found at: build/libs/.

An uber jar containing the library and all its relocated dependencies except the kafka client and slf4j-api can also be built. Use this command:

gradle clean shadowJar

The generated uber jar file can also be found at: build/libs/. At runtime, the uber jar expects to find the kafka client library and the sl4j-api library on the classpath.

Validating secure dependencies

To ensure no security vulnerabilities in the dependency libraries, run the following.

gradle dependencyCheckAnalyze

If the above reports any vulnerabilities, upgrade dependencies to use the respective latest versions.

Using the Amazon MSK Library for IAM Authentication

The recommended way to use this library is to consume it from maven central while building a Kafka client application.

<dependency>
    <groupId>software.amazon.msk</groupId>
    <artifactId>aws-msk-iam-auth</artifactId>
    <version>1.1.5</version>
</dependency>

If you want to use it with a pre-existing Kafka client, you could build the uber jar and place it in the Kafka client's classpath.

Configuring a Kafka client to use AWS IAM

You can configure a Kafka client to use AWS IAM for authentication by adding the following properties to the client's configuration.

# Sets up TLS for encryption and SASL for authN.
security.protocol = SASL_SSL

# Identifies the SASL mechanism to use.
sasl.mechanism = AWS_MSK_IAM

# Binds SASL client implementation.
sasl.jaas.config = software.amazon.msk.auth.iam.IAMLoginModule required;

# Encapsulates constructing a SigV4 signature based on extracted credentials.
# The SASL client bound by "sasl.jaas.config" invokes this class.
sasl.client.callback.handler.class = software.amazon.msk.auth.iam.IAMClientCallbackHandler

This configuration finds IAM credentials using the AWS Default Credentials Provider Chain. To summarize, the Default Credential Provider Chain looks for credentials in this order:

1. Environment variables: AWS_ACCESS_KEY_ID and AWS_SECRET_ACCESS_KEY.
2. Java system properties: aws.accessKeyId and aws.secretKey.
3. Web Identity Token credentials from the environment or container.
4. The default credential profiles file– typically located at ~/.aws/credentials (location can vary per platform), and shared by many of the AWS SDKs and by the AWS CLI.
   You can create a credentials file by using the aws configure command provided by the AWS CLI, or you can create it by editing the file with a text editor. For information about the credentials file format, see AWS Credentials File Format.
5. It can be used to load credentials from credential profiles other than the default one by setting the environment variable
   AWS_PROFILE to the name of the alternate credential profile. Profiles can be used to load credentials from other sources such as AWS IAM Roles. See AWS Credentials File Format for more details.
6. Amazon ECS container credentials– loaded from the Amazon ECS if the environment variable AWS_CONTAINER_CREDENTIALS_RELATIVE_URI is set.
7. Instance profile credentials: used on EC2 instances, and delivered through the Amazon EC2 metadata service.

Specifying an alternate credential profile for a client

If the client wants to specify a particular credential profile as part of the client configuration rather than through the environment variable AWS_PROFILE, they can pass in the name of the profile as a client configuration property:

# Binds SASL client implementation. Uses the specified profile name to look for credentials.
sasl.jaas.config = software.amazon.msk.auth.iam.IAMLoginModule required awsProfileName="<Credential Profile Name>";

Specifying a role based credential profile for a client

Some clients may want to assume a role and use the role's temporary credentials to communicate with a MSK cluster. One way to do that is to create a credential profile for that role following the rules for Using an IAM role in the CLI. They can then pass in the name of the credential profile as described above.

As an example, let's say a Kafka client is running on an Ec2 instance and the Kafka client wants to use an IAM role called msk_client_role. The Ec2 instance profile has permissions to assume the msk_client_role IAM role although msk_client_role is not attached to the instance profile.

In such a case, we create a credential profile called msk_client that assumes the role msk_client_role. The credential profile looks like:

[msk_client]
role_arn = arn:aws:iam::123456789012:role/msk_client_role
credential_source = Ec2InstanceMetadata

The credential profile name msk_client is passed in as a client configuration property:

sasl.jaas.config = software.amazon.msk.auth.iam.IAMLoginModule required awsProfileName="msk_client";

Many more examples of configuring credential profiles with IAM roles can be found in Using an IAM role in the CLI.

Specifying an AWS IAM Role for a client

The library supports another way to configure a client to assume an IAM role and use the role's temporary credentials. The IAM role's ARN and optionally the session name for the client can be passed in as client configuration property:

sasl.jaas.config=software.amazon.msk.auth.iam.IAMLoginModule required awsRoleArn="arn:aws:iam::123456789012:role/msk_client_role" awsRoleSessionName="producer"  awsStsRegion="us-west-2";

In this case, the awsRoleArn specifies the ARN for the IAM role the client should use and awsRoleSessionName specifies the session name that this particular client should use while assuming the IAM role. If the same IAM Role is used in multiple contexts, the session names can be used to differentiate between the different contexts. The awsRoleSessionName is optional.

awsStsRegion optionally specifies the regional endpoint of AWS STS to use while assuming the IAM role. If awsStsRegion is omitted the global endpoint for AWS STS is used by default. When the Kafka client is running in a VPC with an STS interface VPC Endpoint (AWS PrivateLink) to a regional endpoint of AWS STS and we want all STS traffic to go over that endpoint, we should set awsStsRegion to the region corresponding to the interface VPC Endpoint. It may also be necessary to configure the sts_regional_endpoints shared AWS config file setting, or the AWS_STS_REGIONAL_ENDPOINTS environment variable as per the AWS STS Regionalized endpoints documentation.

The Default Credential Provider Chain must contain the permissions necessary to assume the client role. For example, if the client is an EC2 instance, its instance profile should have permission to assume the msk_client_role.

Figuring out whether or not to use default credentials

When you want the MSK client to connect to MSK using credentials not found in the AWS Default Credentials Provider Chain, you can specify an awsProfileName containing the credential profile to use, or an awsRoleArn to indicate an IAM Role’s ARN to assume using credentials in the Default Credential Provider Chain. These parameters are optional, and if they are not set the MSK client will use credentials from the Default Credential Provider Chain. There is no need to specify them if you intend to use an IAM role associated with an AWS compute service, such as EC2 or ECS to authenticate to MSK.

Retries while getting credentials

In some scenarios the IAM credentials might be transiently unavailable. This will cause the connection to fail, which might in some cases cause the client application to stop. So, in version 1.1.3 the library retries loading the credentials when it gets an SkdClientException (which wraps most AWS SDK client side exceptions). Since the retries do not impact the fault-free path and we had heard of user issues around random failures loading credentials (e.g.: #59, maybe #51 ), we decided to change the default behavior to retry a maximum of 3 times. It exponentially backs off with full jitter upto a max-delay of 2000 ms.

The maximum number of retries and the maximum back off period can be set:

sasl.jaas.config=software.amazon.msk.auth.iam.IAMLoginModule required awsMaxRetries="7" awsMaxBackOffTimeMs="500";

This sets the maximum number of retries to 7 and the maximum back off time to 500 ms.

The retries can be turned off completely by setting awsMaxRetries to "0".

Troubleshooting

IAMClientCallbackHandler could not be found

A Kafka client configured to use AWS_MSK_IAM may see an error that the IAMClientCallbackHandler cannot be found:

Exception in thread "main" org.apache.kafka.common.config.ConfigException: Invalid value 
software.amazon.msk.auth.iam.IAMClientCallbackHandler for configuration sasl.client.callback.handler.class: 
Class software.amazon.msk.auth.iam.IAMClientCallbackHandler could not be found.

That means that this aws-msk-iam-auth library is not on the classpath of the Kafka client. Please add the aws-msk-iam-auth library to the classpath and try again.

Finding out which identity is being used

You may receive an Access denied error and there may be some doubt as to which credential is being exactly used. The credential may be sourced from a role ARN, EC2 instance profile, credential profile etc. This may be particularly so when cross account access is being attempted. If the client side logging is set to DEBUG and the client configuration property includes awsDebugCreds set to true:

sasl.jaas.config=software.amazon.msk.auth.iam.IAMLoginModule required awsDebugCreds=true;

the client library will print a debug log of the form:

DEBUG The identity of the credentials is {UserId: ARID4JIC6BCC6OK5OFDFGS:test124,Account: 12345678,Arn: arn:aws:sts::12345678:assumed-role/kada/test124} (software.amazon.msk.auth.iam.internals.MSKCredentialProvider)```

The log line provides the IAM Account, IAM user id and the ARN of the IAM Principal corresponding to the credential being used. The awsDebugCreds=true parameter can be combined with any of the other parameters such as awsRoleArn, awsRoleSessionName.

Please note that the log level should also be set to DEBUG for this information to be logged. It is not recommended to run with awsDebugCreds=true since it makes an additional remote call.

Failed Authentication: Too many connects

You may receive an error, indicating that authentication has failed due to Too many connects, similar to:

 ERROR org.apache.kafka.clients.NetworkClient - [Producer clientId=producer-x] Connection to node 3 (...) failed
 authentication due to: [446c81dc-9ab3-4d4b-b174-4ecd9baa406c]: Too many connects

This is a sign that one or more IAM clients are trying to connect to a particular broker too many times per second and the broker is protecting itself.

Setting the reconnect.backoff.ms to at least 1000 should help clients backoff and retry connections such that the broker does not need to reject new connections because of the connection rate.

The broker type determines the limit on the rate of new IAM connections per broker. Please note the limit is not about the total number of connections per broker but the rate of new IAM connections per broker. See the limits page for MSK for the limit on the rate of new IAM connections per broker for different broker types.

Kafka Connect: Unsupported callback type

While using the library from a Kafka Connect client, you may see an error of the form:

[Producer clientId=connector-...] Failed authentication with BROKER (An error: (java.security.PrivilegedActionException: javax.security.sasl.SaslException: Exception while evaluating challenge [Caused by javax.security.auth.callback.UnsupportedCallbackException: Unsupported callback type:

This most commonly occurs when two different class loaders are used to load different classes used by the library. It can happen when the aws-msk-iam-auth library is placed on the plugin path for Kafka Connect. Since the library is actually used by the Kafka producer and consumer clients and not the Kafka Connect plugin itself, it should be placed in a location that is on the classpath but outside the plugin path. This should ensure that Kafka Connect's PluginClassLoader is not used to load classes for the aws-msk-iam-auth library.

Dependency mismatch

If you are building the library from source using gradle build and copying it over to a Kafka client on that or another machine, there is a chance that some dependencies may not be available on the Kafka client machine. In that case, you could instead generate and use the uber jar that packages all the necessary runtime dependencies by running gradle shadowJar.

Details

This library introduces a new SASL mechanism called AWS_MSK_IAM. The IAMLoginModule is used to register the IAMSaslClientProvider as a Provider for the AWS_MSK_IAM mechanism. The IAMSaslClientProvider is used to generate a new IAMSaslClient every time a new connection to a Kafka broker is opened or an existing connection is re-authenticated.

The IAMSaslClient is used to perform the actual SASL authentication for a client. It evaluates challenges and creates responses that can be used to authenticate a Kafka client to a Kafka broker configured with AWS IAM authentication . Its initial response contains an authentication payload that includes a signature generated using the client's credentials. The IAMClientCallbackHandler is used to extract the client credentials that are then used for generating the signature.

The authentication payload and the signature are generated by the AWS4SignedPayloadGenerator class based on the parameters specified in AuthenticationRequestParams. The authentication payload consists of a JSON object:

{
    "version" : "2020_10_22",
    "host" : "<broker address>",
    "user-agent": "<user agent string from the client>",
    "action": "kafka-cluster:Connect",
    "x-amz-algorithm" : "<algorithm>",
    "x-amz-credential" : "<clientAWSAccessKeyID>/<date in yyyyMMdd format>/<region>/kafka-cluster/aws4_request",
    "x-amz-date" : "<timestamp in yyyyMMdd'T'HHmmss'Z' format>",
    "x-amz-security-token" : "<clientAWSSessionToken if any>",
    "x-amz-signedheaders" : "host",
    "x-amz-expires" : "<expiration in seconds>",
    "x-amz-signature" : "<AWS SigV4 signature computed by the client>"
}

Generating the authentication payload

Please note that all the keys in the authentication payload are in lowercase.

The values of the following keys in the authentication payload are fixed for a client:

• "version" currently always has the value "2020_10_22"
• "user-agent" is a string passed in by the client library to describe the client. The simplest user agent is "<name of client library>". However, more details can be added to the user agent as well "<name of client library /<version of client library>/<os and version>/<version of language>"

The values of the remaining keys will be generated as part of calculating the signature of the authentication payload . The signature is calculated by following the rules for generating presigned urls. Although, this library uses the AWS4Signer from the AWS SDK for Java to generate the signature we outline the steps followed in calculating the signature and generating the authentication payload from it.

The inputs to this calculation are

1. The AWS Credentials that will be used to sign the authentication payload. This has 3 parts: the client AWSAccessKeyId, the client AWSSecretyKeyId and the optional client SessionToken.
2. The hostname of the Kafka broker to which the client wishes to connect.
3. The AWS region in which the Kafka broker exists.
4. The timestamp when the connection is being established.

The steps in the calculation are:

1. Generate a canonical request based on the inputs.
2. Generate a string to sign based on the canonical request.
3. Calculate the signature based on the string to sign and the inputs.
4. Put it all together in the authentication payload.

Generate a Canonical Request

We start by generating a canonical request with an empty payload based on the inputs. The canonical request in this case has the following form

"GET\n"+
"/\n"+
<CanonicalQueryString>+"\n"+
<CanonicalHeaders>+"\n"+
<SignedHeaders>+"\n"+
<HashedPayload>

Canonical Query String

<CanonicalQueryString> specifies the authentication parameters as URI-encoded query string parameters. We URI-encode query parameter names and values individually. We also sort the parameters in the canonical query string alphabetically by key name. The sorting occurs after encoding. The <CanonicalQueryString> can be calculated by:

UriEncode("Action")+"="+UriEncode("kafka-cluster:Connect")+"&"+
UriEncode("X-Amz-Algorithm")+"="+UriEncode("AWS4-HMAC-SHA256") + "&" +
UriEncode("X-Amz-Credential")+"="+UriEncode("<clientAWSAccessKeyID>/<timestamp in yyyyMMdd format>/<AWS region>/kafka-cluster/aws4_request") + "&" +
UriEncode("X-Amz-Date")+"="+UriEncode("<timestamp in yyyyMMdd'T'HHmmss'Z' format>") + "&" +
UriEncode("X-Amz-Expires")+"="+UriEncode("900") + "&" +
UriEncode("X-Amz-Security-Token")+"="+UriEncode("<client Session Token>") + "&" +
UriEncode("X-Amz-SignedHeaders")+"="+UriEncode("host")

The exact definition of URIEncode from generating presigned urls maybe found later.

The query string parameters are in order:

• "Action": Always has the value "kafka-cluster:Connect"
• "X-Amz-Algorithm": Describes the algorithm used to calculate the signature. Currently it is "AWS4-HMAC-SHA256"
• "X-Amz-Credential": Contains the access key ID, timestamp in yyyyMMdd format, the scope of the credential and the constant string aws4_request. The scope is defined as the AWS region of the Kafka broker and the name of the service ("kafka-cluster" in this case). For example if the broker is in us-west-2 region, the scope is us-west-2/kafka-cluster. This scope will be used again later to calculate the signature in String To Sign and must match the one used to calculate the signature.
• "X-Amz-Date": The date and time format must follow the ISO 8601 standard, and must be formatted with the "yyyyMMddTHHmmssZ" format. The date and time must be in UTC.
• "X-Amz-Expires" : Provides the time period, in seconds, for which the generated presigned URL is valid. We recommend 900 seconds.
• "X-Amz-Security-Token": The session token if it is specified as part of the AWS Credential. Otherwise this query parameter is skipped.
• "X-Amz-Signedheaders" is currently always set to host.

Canonical Header

<CanonicalHeaders> is a list of request headers with their values. Header names must be in lowercase. Individual header name and value pairs are separated by the newline character ("\n"). In this case there is just one header. So <CanonicalHeaders> is calculated by:

"host"+":"+"<broker host name>"+"\n"

Signed Headers

<SignedHeaders> is an alphabetically sorted, semicolon-separated list of lowercase request header names. In this case there is just one header. So <SignedHeaders> is calculated by:

"host"

Hashed Payload

Since the payload is empty the <HashedPayload> is calculated as

Hex(SHA256Hash(""))

where

• Hex is a function to do lowercase base 16 encoding.
• SHA256Hash is a Secure Hash Algorithm (SHA) cryptographic hash function.

String To Sign

From the canonical string, we derive the string that will be used to sign the authentication payload. The String to Sign is calculated as:

"AWS4-HMAC-SHA256" + "\n" +
"<timestamp in yyyyMMdd format>" + "\n" +
<Scope> + "\n" +
Hex(SHA256Hash(<CanonicalRequest>))

where

• Hex is a function to do lowercase base 16 encoding.
• SHA256Hash is a Secure Hash Algorithm (SHA) cryptographic hash function.

The <Scope> is defined as the AWS region of the Kafka broker and the name of the service ("kafka-cluster" in this case). For example if the broker is in us-west-2 region, the scope is "us-west-2/kafka-cluster". It must be the same scope as was defined for the "X-Amz-Credential" query parameter while generating the canonical query string.

Calculate Signature

The signature is calculated by:

DateKey              = HMAC-SHA256("AWS4"+"<client AWSSecretAccessKey>", "<timestanp in YYYYMMDD>")
DateRegionKey        = HMAC-SHA256(<DateKey>, "<aws-region>")
DateRegionServiceKey = HMAC-SHA256(<DateRegionKey>, "kafka-cluster")
SigningKey           = HMAC-SHA256(<DateRegionServiceKey>, "aws4_request")
Signature            = Hex(HMAC-SHA256(<SigningKey>, <StringToSign>))

where

• Hex is a function to do lowercase base 16 encoding.
• HMAC-SHA256 is a function that computes HMAC by using the SHA256 algorithm with the signing key provided.

The <Signature> is the final signature.

Putting it all together

As mentioned earlier, the authentication payload is a json object with certain keys. All the keys are in lower case. The following keys in the authentication payload json are constant:

• "version"
• "user-agent"

All the query parameters calculated earlier are added to the authentication payload json. The keys are lower case strings and the values are the ones calculated earlier but the values are not uri encoded:

• "action"
• "x-amz-algorithm"
• "x-amz-credential"
• "x-amz-date"
• "x-amz-expires"
• "x-amz-security-token"
• "x-amz-signedheaders"

The host header is added to the authentication payload json

• "host" and its value is set to the hostname of the broker being connected.

The <Signature> calculated in the previous step is added to the authentication payload json as

• "x-amz-signature" and its value is set to<Signature>

This finally yields the authentication payload that looks like

{
    "version" : "2020_10_22",
    "host" : "<broker address>",
    "user-agent": "<user agent string from the client>",
    "action": "kafka-cluster:Connect",
    "x-amz-algorithm" : "<algorithm>",
    "x-amz-credential" : "<clientAWSAccessKeyID>/<date in yyyyMMdd format>/<region>/kafka-cluster/aws4_request",
    "x-amz-date" : "<timestamp in yyyyMMdd'T'HHmmss'Z' format>",
    "x-amz-security-token" : "<clientAWSSessionToken if any>",
    "x-amz-signedheaders" : "host",
    "x-amz-expires" : "<expiration in seconds>",
    "x-amz-signature" : "<AWS SigV4 signature computed by the client>"
}

Message Exchange with Kafka Broker

This authentication payload is sent as the first message from the client to the Kafka broker. The kafka broker then responds with a challenge. We expect a non-empty response from the broker if authentication using AWS IAM succeeded. The authentication response is a json object that may be logged:

{
  "version" : "2020_10_22",
  "request-id" : "<broker generated request id>"
}

The request-id which is generated by the broker can be useful for debugging issues with AWS IAM authentication between the client and the broker.

UriEncode

Snipped from the detailed rules for generating presigned urls. URI encode every byte. UriEncode() must enforce the following rules:

• URI encode every byte except the unreserved characters: 'A'-'Z', 'a'-'z', '0'-'9', '-', '.', '_', and '~'.
• The space character is a reserved character and must be encoded as "%20" (and not as "+").
• Each URI encoded byte is formed by a '%' and the two-digit hexadecimal value of the byte.
• Letters in the hexadecimal value must be uppercase, for example "%1A".
• Encode the forward slash character, '/', everywhere except in the object key name. For example, if the object key name is photos/Jan/sample.jpg, the forward slash in the key name is not encoded.

The following is an example UriEncode() function in Java.

public class UriEncode {
public static String UriEncode(CharSequence input, boolean encodeSlash) {
          StringBuilder result = new StringBuilder();
          for (int i = 0; i < input.length(); i++) {
              char ch = input.charAt(i);
              if ((ch >= 'A' && ch <= 'Z') || (ch >= 'a' && ch <= 'z') || (ch >= '0' && ch <= '9') || ch == '_' || ch == '-' || ch == '~' || ch == '.') {
                  result.append(ch);
              } else if (ch == '/') {
                  result.append(encodeSlash ? "%2F" : ch);
              } else {
                  result.append(toHexUTF8(ch));
              }
          }
          return result.toString();
      } 
}

Release Notes

Release 1.1.5

• Update dependencies to address the following security vulnerabilities.
  • CVE-2022-42003
  • CVE-2022-42004
• Add support for multi-classloader environments, such as Apache Flink (#36)

Release 1.1.4

• Update dependencies to address the following security vulnerabilities.
  • CVE-2021-37136
  • CVE-2021-37137
  • CVE-2022-24823
  • CVE-2021-43797
  • CVE-2021-38153
  • CVE-2020-36518
• Specifically, build and test against Kafka 2.8.

Release 1.1.3

• Add retries if loading credential fails with client side errors.
• If AWS STS is not accessible for identifying the credential when awsDebugCreds=true, do not fail the connection.
• Update Troubleshooting section in README.

Release 1.1.2

• Update log4j version in test dependencies to CVE-2021-44832
• Allow users to debug the credentials being used

Release 1.1.1

• Enable support for STS regional endpoints when configured to assume a role (thanks dvuple@)
• Additional logging to log the classes and classloaders for IAMClientCallbackHandler and AWSCredentialsCallback classes.
• README updates to start a section on troubleshooting.
• In the uber jar do not relocate the awsk sdk v2 modules.
• In the uber jar, stop shadowing slf4j-api.

Release 1.1.0

• Add support for credential profiles based on AWS Single Sign-On (SSO).
• Add support for clients using IAM Roles without using credential profiles.
• Bug fix for credential profiles with IAM Roles.

Release 1.0.0

• First version of the Amazon MSK Library for IAM Authentication

See CONTRIBUTING for more information.