Api Configuration¶

A domain node exposes two main APIs: the admin-api and the public-api, while the participant node exposes the ledger-api and the admin-api. In this section, we will explain what the APIs do and how they can be configured. All APIs are based on GRPC, which is an efficient RPC and streaming protocol with client support in almost all relevant programming languages. Native bindings can be built using the API definitions.

Default Ports¶

Canton assigns ports automatically for all the APIs of all the configured nodes if the port has not been configured explicitly. The ports are allocated according to the following scheme:

/** Participant node default ports */
val ledgerApiPort = defaultPortStart(4001)
val participantAdminApiPort = defaultPortStart(4002)

/** Domain node default ports */
val domainPublicApiPort = defaultPortStart(4201)
val domainAdminApiPort = defaultPortStart(4202)

/** External sequencer node default ports (enterprise-only) */
val sequencerPublicApiPort = defaultPortStart(4401)
val sequencerAdminApiPort = defaultPortStart(4402)

/** External mediator node default port (enterprise-only) */
val mediatorAdminApiPort = defaultPortStart(4602)

/** Domain node default ports */
val domainManagerAdminApiPort = defaultPortStart(4801)

/** Increase the default port number for each new instance by portStep */
private val portStep = 10

Administration API¶

The nature and scope of the admin api on participant and domain nodes has some overlap. As an example, you will find the same key management commands on the domain and the participant node API, whereas the participant has different commands to connect to several domains.

The configuration currently is simple (see the TLS example below) and just takes an address and a port. The address defaults to 127.0.0.1 and a default port is assigned if not explicitly configured.

You should not expose the admin-api publicly in an unsecured way as it serves administrative purposes only.

TLS Configuration¶

Both, the Ledger API and the admin API provide the same TLS capabilities and can be configured using the same configuration directives. TLS provides end-to-end channel encryption between the server and client, and depending on the settings, server or mutual authentication.

A full configuration example is given by

canton.participants.participant4.ledger-api {
  address = "127.0.0.1" // IP / DNS must be SAN of certificate to allow local connections from the canton process
  port = 5041
  tls {
    // the certificate to be used by the server
    cert-chain-file = "./tls/participant.crt"
    // private key of the server
    private-key-file = "./tls/participant.pem"
    // trust collection, which means that all client certificates will be verified using the trusted
    // certificates in this store. if omitted, the JVM default trust store is used.
    trust-collection-file = "./tls/root-ca.crt"
    // define whether clients need to authenticate as well (default not)
    client-auth = {
      // none, optional and require are supported
      type = require
      // If clients are required to authenticate as well, we need to provide a client
      // certificate and the key, as Canton has internal processes that need to connect to these
      // APIs. If the server certificate is trusted by the trust-collection, then you can
      // just use the server certificates. Otherwise, you need to create separate ones.
      admin-client {
        cert-chain-file = "./tls/admin-client.crt"
        private-key-file = "./tls/admin-client.pem"
      }
    }
    // minimum-server-protocol-version = ...
    // ciphers = ...
  }
}

These TLS settings allow a connecting client to ensure that it is talking to the right server. In this example, we have also enabled client authentication, which means that the client needs to present a valid certificate (and have the corresponding private key). The certificate is valid if it has been signed by a key in the trust store.

The trust-collection-file allows us to provide a file based trust store. If omitted, the system will default to the built-in JVM trust store. The file must contain all client certificates (or parent certificates which were used to sign the client certificate) who are trusted to use the API. The format is just a collection of PEM certificates (in the right order or hierarchy), not a java based trust store.

In order to operate the server just with server-side authentication, you can just omit the section on client-auth. However, if client-auth is set to require, then Canton also requires a client certificate, as various Canton internal processes will connect to the process itself through the API.

All the private keys need to be in the pkcs8 PEM format.

By default, Canton only uses new versions of TLS and strong ciphers. You can also override the default settings using the variables ciphers and protocols. If you set these settings to null, the default JVM values will be used.

Note

Error messages on TLS issues provided by the networking library netty are less than optimal. If you are struggling with setting up TLS, please enable DEBUG logging on the io.netty logger.

Note that the configuration hierarchy for a remote participant console is slightly different from the in-process console or participant shown above. For configuring a remote console with TLS, please see the scaladocs for a TlsClientConfig (see also how scaladocs relates to the configuration).

If you need to create a set of testing TLS certificates, you can use the following openssl commands:

DAYS=3650

function create_key {
  local name=$1
  openssl genrsa -out "${name}.key" 4096
  # netty requires the keys in pkcs8 format, therefore convert them appropriately
  openssl pkcs8 -topk8 -nocrypt -in "${name}.key" -out "${name}.pem"
}

# create self signed certificate
function create_certificate {
  local name=$1
  local subj=$2
  openssl req -new -x509 -sha256 -key "${name}.key" \
              -out "${name}.crt" -days ${DAYS} -subj "$subj"
}

# create certificate signing request with subject and SAN
# we need the SANs as our certificates also need to include localhost or the
# loopback IP for the console access to the admin-api and the ledger-api
function create_csr {
  local name=$1
  local subj=$2
  local san=$3
  (
    echo "authorityKeyIdentifier=keyid,issuer"
    echo "basicConstraints=CA:FALSE"
    echo "keyUsage = digitalSignature, nonRepudiation, keyEncipherment, dataEncipherment"
  ) > ${name}.ext
  if [[ -n $san ]]; then
    echo "subjectAltName=${san}" >> ${name}.ext
  fi
  # create certificate (but ensure that localhost is there as SAN as otherwise, admin local connections won't work)
  openssl req -new -sha256 -key "${name}.key" -out "${name}.csr" -subj "$subj"
}

function sign_csr {
  local name=$1
  local sign=$2
  openssl x509 -req -sha256 -in "${name}.csr" -extfile "${name}.ext" -CA "${sign}.crt" -CAkey "${sign}.key" -CAcreateserial  \
               -out "${name}.crt" -days ${DAYS}
  rm "${name}.ext" "${name}.csr"
}

function print_certificate {
  local name=$1
  openssl x509 -in "${name}.crt" -text -noout
}

# create root certificate
create_key "root-ca"
create_certificate "root-ca" "/O=TESTING/OU=ROOT CA/emailAddress=canton@digitalasset.com"
#print_certificate "root-ca"

# create domain certificate
create_key "domain"
create_csr "domain" "/O=TESTING/OU=DOMAIN/CN=localhost/emailAddress=canton@digitalasset.com" "DNS:localhost,IP:127.0.0.1"
sign_csr "domain" "root-ca"
print_certificate "domain"

# create participant certificate
create_key "participant"
create_csr "participant" "/O=TESTING/OU=PARTICIPANT/CN=localhost/emailAddress=canton@digitalasset.com" "DNS:localhost,IP:127.0.0.1"
sign_csr "participant" "root-ca"

# create participant client key and certificate
create_key "admin-client"
create_csr "admin-client" "/O=TESTING/OU=ADMIN CLIENT/CN=localhost/emailAddress=canton@digitalasset.com"
sign_csr "admin-client" "root-ca"
print_certificate "admin-client"

Keep Alive¶

In order to prevent load-balancers or firewalls from terminating long running RPC calls in the event of some silence on the connection, all GRPC connections enable keep-alive by default. An example configuration for an adjusted setting is given below:

canton.participants.participant2 {
  admin-api         {
    address = "127.0.0.1"
    port = 5022
    keep-alive-server {
      time = 40s
      timeout = 20s
      permit-keep-alive-time = 20s
    }
  }
  sequencer-client {
    keep-alive-client {
      time = 60s
      timeout = 30s
    }
  }
}

GRPC client connections are configured with keep-alive-client, with two settings: time, and timeout. The effect of the time and timeout settings are described in the GRPC documentation.

Servers can additionally change another setting: permit-keep-alive-time. This specifies the most aggressive keep-alive time that a client is permitted to use. If a client uses keep-alive time that is more aggressive than the permit-keep-alive-time, the connection will be terminated with a GOAWAY frame with “too_many_pings” as the debug data. This setting is described in more detail in the GRPC documentation and GRPC manual page.

Max Inbound Message Size¶

The APIs exposed by both the participant (ledger API and admin API) as well as by the domain (public API and admin API) have an upper limit on incoming message size. To increase this limit to accommodate larger payloads, the flag max-inbound-message-size has to be set for the respective API to the maximum message size in bytes.

For example, to configure a participant’s ledger API limit to 20MB:

canton.participants.participant2.ledger-api {
  address = "127.0.0.1"
  port = 5021
  max-inbound-message-size = 20971520
}

Participant Configuration¶

Ledger Api¶

The configuration of the ledger API is similar to the admin API configuration, except that the group starts with ledger-api instead of admin-api.

JWT Authorization¶

The Ledger Api supports JWT based authorization checks. Please consult the Daml SDK manual to understand the various configuration options and their security aspects. Canton exposes precisely the same JWT authorization options as explained therein.

In order to enable JWT authorization checks, your safe configuration options are

_shared {
  ledger-api {
    auth-services = [{
      // type can be
      //   jwt-rs-256-crt
      //   jwt-es-256-crt
      //   jwt-es-512-crt
      type = jwt-rs-256-crt
      // we need a certificate file (abcd.cert)
      certificate = ${JWT_CERTIFICATE_FILE}
    }]
  }
}

_shared {
  ledger-api {
    auth-services = [{
      type = jwt-rs-256-jwks
      // we need a URL to a jwks key, e.g. https://path.to/jwks.key
      url = ${JWT_URL}
    }]
  }
}

while there is also unsafe HMAC256 based support, which can be enabled using

_shared {
  ledger-api {
    auth-services = [{
      type = unsafe-jwt-hmac-256
      secret = "not-safe-for-production"
    }]
  }
}

Note that you can define multiple authorization plugins. If more than one is defined, the system will use the claim of the first auth plugin that does not return Unauthorized.

Leeway Parameters for JWT Authorization¶

You can define leeway parameters for authorization using JWT tokens. An authorization which fails due to clock skew between the signing and the verification of the tokens can be eased by specifying a leeway window in which the token should still be considered valid. Leeway can be defined either specifically for the Expiration Time (“exp”), Not Before (“nbf”) and Issued At (“iat”) claims of the token or by a default value for all three. The values defining the leeway for each of the three specific fields override the default value if present. The leeway parameters should be given in seconds and can be defined as in the example configuration below:

_shared {
  parameters.ledger-api-server-parameters.jwt-timestamp-leeway {
    default = 5
    expires-at = 10
    issued-at = 15
    not-before = 20
  }
}

Limiting concurrent GRPC requests¶

(Preview feature) The Ledger Api server limits the number of parallel requests accepted on the Api based on system resource usage. We recommend to keep the default values. However, they can be changed using the following configuration options:

// re-configure the default rate limit values
canton.participants.participant1.ledger-api.rate-limit = {
    max-api-services-queue-size = 20000
    max-streams = 1000
    // These settings are highly sensitive to the operating environment and should
    // only be configured where memory profiling has highlighted spikes in memory
    // usage that need to be flattened
    max-used-heap-space-percentage = 100
    min-free-heap-space-bytes = 0
}
// turn off rate limits for participant2
canton.participants.participant2.ledger-api.rate-limit = null

Turning the rate limits off can be achieved by setting them to null.

Domain Configurations¶

Public Api¶

The domain configuration requires the same configuration of the admin-api as the participant. Next to the admin-api, we need to configure the public-api, which is the api where all participants connect.

Authentication Token¶

Authentication of the restricted services is built into the public sequencer api, leveraging the participant signing keys. You don’t need to do anything in order to set this up; it is enforced automatically and can’t be turned off. The same mechanism is used to check the authentication of the domain topology manager and the mediator.

The token is generated during the handshake between the node and the sequencer. By default, it is valid for one hour. The nodes automatically renew the token in the background before it expires. The lifetime of the tokens and of the nonce can be reconfigured using

canton.domains.mydomain.public-api {
    token-expiration-time = 60m
    nonce-expiration-time = 1m
}

However, we suggest keeping the default values.

TLS Encryption¶

As with the admin-api, network traffic can (and should) be encrypted using TLS. This is particularly crucial for the Public API.

An example configuration section which enables TLS encryption and server-side TLS authentication is given by

canton.domains.acme.public-api {
  port = 5028
  address = localhost // defaults to 127.0.0.1
  tls {
    cert-chain-file = "./tls/domain.crt"
    private-key-file = "./tls/domain.pem"
    // minimum-server-protocol-version = TLSv1.3, optional argument
    // ciphers = null // use null to default to JVM ciphers
  }
}

If TLS is used on the server side with a self-signed certificate, we need to pass the certificate chain during the connect call of the participant. Otherwise, the default root certificates of the Java runtime will be used. An example would be:

participant3.domains.connect(
  domainAlias = "acme",
  connection = s"https://$hostname:$port",
  certificatesPath = certs, // path to certificate chain file (.pem) of server
)

Limiting concurrent GRPC requests (preview feature)¶

In large-scale deployments a Canton node may get more GRPC requests than it can cope with, leading to requests timing out. Canton has an experimental integration with concurrency-limits to limit the number of concurrent requests and prevent nodes from becoming overloaded.

Canton’s GRPC services can be configured to use various options provided by the concurrency-limits library:

A fixed limit on concurrent requests

canton.participants.participant1.admin-api.concurrency-limit {
    type = "fixed-limit"
    limit = "10"
}

A dynamic limit for the number of concurrent requests, inspired by TCP Vegas

canton.participants.participant1.admin-api.concurrency-limit {
    # Values are defaults from https://github.com/Netflix/concurrency-limits
    type = "vegas-limit"
    max-concurrency = 1000
    smoothing = 1.0
}

A gradient-based algorithm to dynamically infer the concurrency limit

canton.participants.participant1.admin-api.concurrency-limit {
    # Values are defaults from https://github.com/Netflix/concurrency-limits
    type = "gradient-2-limit"
    max-concurrency = 200
    smoothing = 0.2
    rtt-tolerance = 1.5
}

Any of these options, with an added smoothing window

canton.participants.participant1.admin-api.concurrency-limit {
    # Values are defaults from https://github.com/Netflix/concurrency-limits
    type = "windowed-limit"
    window-size = 10
    delegate = {
        type = gradient-2-limit
        max-concurrency = 200
        smoothing = 0.2
        rtt-tolerance = 1.5
    }
}

See the concurrency-limits library for more information on these different options.