corda/docs/source/network-map.rst

The network map

The network map is a collection of signed NodeInfo objects. Each NodeInfo is signed by the node it represents and thus cannot be tampered with. It forms the set of reachable nodes in a compatibility zone. A node can receive these objects from two sources:

1. A network map server that speaks a simple HTTP based protocol.
2. The additional-node-infos directory within the node's directory.

The network map server also distributes the parameters file that define values for various settings that all nodes need to agree on to remain in sync.

Note

In Corda 3 no implementation of the HTTP network map server is provided. This is because the details of how a compatibility zone manages its membership (the databases, ticketing workflows, HSM hardware etc) is expected to vary between operators, so we provide a simple REST based protocol for uploading/downloading NodeInfos and managing network parameters. A future version of Corda may provide a simple "stub" implementation for running test zones. In Corda 3 the right way to run a test network is through distribution of the relevant files via your own mechanisms. We provide a tool to automate the bulk of this task (see below).

HTTP network map protocol

If the node is configured with the compatibilityZoneURL config then it first uploads its own signed NodeInfo to the server at that URL (and each time it changes on startup) and then proceeds to download the entire network map from the same server. The network map consists of a list of NodeInfo hashes. The node periodically polls for the network map (based on the HTTP cache expiry header) and any new entries are downloaded and cached. Entries which no longer exist are deleted from the node's cache.

The set of REST end-points for the network map service are as follows.

Request method	Path	Description
POST	/network-map/publish	For the node to upload its signed NodeInfo object to the network map.
POST	/network-map/ack-parameters	For the node operator to acknowledge network map that new parameters were accepted for future update.
GET	/network-map	Retrieve the current signed public network map object. The entire object is signed with the network map certificate which is also attached.
GET	/network-map/{uuid}	Retrieve the current signed private network map object with given uuid. Format is the same as for /network-map endpoint.
GET	/network-map/node-info/{hash}	Retrieve a signed NodeInfo as specified in the network map object.
GET	/network-map/network-parameters/{hash}	Retrieve the signed network parameters (see below). The entire object is signed with the network map certificate which is also attached.
GET	/network-map/my-hostname	Retrieve the IP address of the caller (and not of the network map).

HTTP is used for the network map service instead of Corda's own AMQP based peer to peer messaging protocol to enable the server to be placed behind caching content delivery networks like Cloudflare, Akamai, Amazon Cloudfront and so on. By using industrial HTTP cache networks the map server can be shielded from DoS attacks more effectively. Additionally, for the case of distributing small files that rarely change, HTTP is a well understood and optimised protocol. Corda's own protocol is designed for complex multi-way conversations between authenticated identities using signed binary messages separated into parallel and nested flows, which isn't necessary for network map distribution.

The additional-node-infos directory

Alongside the HTTP network map service, or as a replacement if the node isn't connected to one, the node polls the contents of the additional-node-infos directory located in its base directory. Each file is expected to be the same signed NodeInfo object that the network map service vends. These are automatically added to the node's cache and can be used to supplement or replace the HTTP network map. If the same node is advertised through both mechanisms then the latest one is taken.

On startup the node generates its own signed node info file, filename of the format nodeInfo-${hash}. It can also be generated using the generate-node-info sub-command without starting the node. To create a simple network without the HTTP network map service simply place this file in the additional-node-infos directory of every node that's part of this network. For example, a simple way to do this is to use rsync.

Usually, test networks have a structure that is known ahead of time. For the creation of such networks we provide a network-bootstrapper tool. This tool pre-generates node configuration directories if given the IP addresses/domain names of each machine in the network. The generated node directories contain the NodeInfos for every other node on the network, along with the network parameters file and identity certificates. Generated nodes do not need to all be online at once - an offline node that isn't being interacted with doesn't impact the network in any way. So a test cluster generated like this can be sized for the maximum size you may need, and then scaled up and down as necessary.

More information can be found in network-bootstrapper.

Network parameters

Network parameters are a set of values that every node participating in the zone needs to agree on and use to correctly interoperate with each other. They can be thought of as an encapsulation of all aspects of a Corda deployment on which reasonable people may disagree. Whilst other blockchain/DLT systems typically require a source code fork to alter various constants (like the total number of coins in a cryptocurrency, port numbers to use etc), in Corda we have refactored these sorts of decisions out into a separate file and allow "zone operators" to make decisions about them. The operator signs a data structure that contains the values and they are distributed along with the network map. Tools are provided to gain user opt-in consent to a new version of the parameters and ensure everyone switches to them at the same time.

If the node is using the HTTP network map service then on first startup it will download the signed network parameters, cache it in a network-parameters file and apply them on the node.

Warning

If the network-parameters file is changed and no longer matches what the network map service is advertising then the node will automatically shutdown. Resolution to this is to delete the incorrect file and restart the node so that the parameters can be downloaded again.

If the node isn't using a HTTP network map service then it's expected the signed file is provided by some other means. For such a scenario there is the network bootstrapper tool which in addition to generating the network parameters file also distributes the node info files to the node directories.

The current set of network parameters:

minimumPlatformVersion

The minimum platform version that the nodes must be running. Any node which is below this will not start.

notaries

List of identity and validation type (either validating or non-validating) of the notaries which are permitted in the compatibility zone.

maxMessageSize

Maximum allowed size in bytes of an individual message sent over the wire. Note that attachments are a special case and may be fragmented for streaming transfer, however, an individual transaction or flow message may not be larger than this value.

maxTransactionSize

Maximum allowed size in bytes of a transaction. This is the size of the transaction object and its attachments.

modifiedTime

The time when the network parameters were last modified by the compatibility zone operator.

epoch

Version number of the network parameters. Starting from 1, this will always increment whenever any of the parameters change.

whitelistedContractImplementations

List of whitelisted versions of contract code. For each contract class there is a list of SHA-256 hashes of the approved CorDapp jar versions containing that contract. Read more about Zone constraints here api-contract-constraints

eventHorizon

Time after which nodes are considered to be unresponsive and removed from network map. Nodes republish their NodeInfo on a regular interval. Network map treats that as a heartbeat from the node.

packageOwnership

List of the network-wide java packages that were successfully claimed by their owners. Any CorDapp JAR that offers contracts and states in any of these packages must be signed by the owner. This ensures that when a node encounters an owned contract it can uniquely identify it and knows that all other nodes can do the same. Encountering an owned contract in a JAR that is not signed by the rightful owner is most likely a sign of malicious behaviour, and should be reported. The transaction verification logic will throw an exception when this happens. Read more about Package ownership here design/data-model-upgrades/package-namespace-ownership.

More parameters will be added in future releases to regulate things like allowed port numbers, whether or not IPv6 connectivity is required for zone members, required cryptographic algorithms and roll-out schedules (e.g. for moving to post quantum cryptography), parameters related to SGX and so on.

Network parameters update process

Network parameters are controlled by the zone operator of the Corda network that you are a member of. Occasionally, they may need to change these parameters. There are many reasons that can lead to this decision: adding a notary, setting new fields that were added to enable smooth network interoperability, or a change of the existing compatibility constants is required, for example.

Note

A future release may support the notion of phased roll-out of network parameter changes.

Updating of the parameters by the zone operator is done in two phases: 1. Advertise the proposed network parameter update to the entire network. 2. Switching the network onto the new parameters - also known as a flag day.

The proposed parameter update will include, along with the new parameters, a human-readable description of the changes as well as the deadline for accepting the update. The acceptance deadline marks the date and time that the zone operator intends to switch the entire network onto the new parameters. This will be a reasonable amount of time in the future, giving the node operators time to inspect, discuss and accept the parameters.

The fact a new set of parameters is being advertised shows up in the node logs with the message "Downloaded new network parameters", and programs connected via RPC can receive ParametersUpdateInfo by using the CordaRPCOps.networkParametersFeed method. Typically a zone operator would also email node operators to let them know about the details of the impending change, along with the justification, how to object, deadlines and so on.

../../core/src/main/kotlin/net/corda/core/messaging/CordaRPCOps.kt

Auto Acceptance

If the only changes between the current and new parameters are for auto-acceptable parameters then, unless configured otherwise, the new parameters will be accepted without user input. The following parameters with the @AutoAcceptable annotation are auto-acceptable:

../../core/src/main/kotlin/net/corda/core/node/NetworkParameters.kt

This behaviour can be turned off by setting the optional node configuration property NetworkParameterAcceptanceSettings.autoAcceptEnabled to false. For example:

...
NetworkParameterAcceptanceSettings {
    autoAcceptEnabled = false
}
...

It is also possible to switch off this behaviour at a more granular parameter level. This can be achieved by specifying the set of @AutoAcceptable parameters that should not be auto-acceptable in the optional NetworkParameterAcceptanceSettings.excludedAutoAcceptableParameters node configuration property.

For example, auto-acceptance can be switched off for any updates that change the packageOwnership map by adding the following to the node configuration:

...
NetworkParameterAcceptanceSettings {
    excludedAutoAcceptableParameters: ["packageOwnership"]
}
...

Manual Acceptance

If the auto-acceptance behaviour is turned off via the configuration or the network parameters change involves parameters that are not auto-acceptable then manual approval is required.

In this case the node administrator can review the change and decide if they are going to accept it. The approval should be done before the update Deadline. Nodes that don't approve before the deadline will likely be removed from the network map by the zone operator, but that is a decision that is left to the operator's discretion. For example the operator might also choose to change the deadline instead.

If the network operator starts advertising a different set of new parameters then that new set overrides the previous set. Only the latest update can be accepted.

To send back parameters approval to the zone operator, the RPC method fun acceptNewNetworkParameters(parametersHash: SecureHash) has to be called with parametersHash from the update. Note that approval cannot be undone. You can do this via the Corda shell (see shell):

run acceptNewNetworkParameters parametersHash: "ba19fc1b9e9c1c7cbea712efda5f78b53ae4e5d123c89d02c9da44ec50e9c17d"

If the administrator does not accept the update then next time the node polls network map after the deadline, the advertised network parameters will be the updated ones. The previous set of parameters will no longer be valid. At this point the node will automatically shutdown and will require the node operator to bring it back again.

Cleaning the network map cache

Sometimes it may happen that the node ends up with an inconsistent view of the network. This can occur due to changes in deployment leading to stale data in the database, different data distribution time and mistakes in configuration. For these unlikely events both RPC method and command line option for clearing local network map cache database exist. To use them you either need to run from the command line:

java -jar corda.jar clear-network-cache

or call RPC method clearNetworkMapCache (it can be invoked through the node's shell as run clearNetworkMapCache, for more information on how to log into node's shell see shell). As we are testing and hardening the implementation this step shouldn't be required. After cleaning the cache, network map data is restored on the next poll from the server or filesystem.