nvflare.private.fed.server.admin module

class ClientReply(client_token: str, req: Message, reply: Message)

Bases: object

Client reply.

Parameters

• client_token (str) – client token

• req (Message) – request

• reply (Message) – reply

class FedAdminServer(fed_admin_interface, users, cmd_modules, file_upload_dir, file_download_dir, allowed_shell_cmds, host, port, ca_cert_file_name, server_cert_file_name, server_key_file_name, accepted_client_cns=None, app_validator=None, download_job_url=None)

Bases: AdminServer

The FedAdminServer is the framework for developing admin commands.

Parameters

• fed_admin_interface – the server’s federated admin interface

• users – a dict of {username: pwd hash}

• cmd_modules – a list of CommandModules

• file_upload_dir – the directory for uploaded files

• file_download_dir – the directory for files to be downloaded

• allowed_shell_cmds – list of shell commands allowed. If not specified, all allowed.

• host – the IP address of the admin server

• port – port number of admin server

• ca_cert_file_name – the root CA’s cert file name

• server_cert_file_name – server’s cert, signed by the CA

• server_key_file_name – server’s private key file

• accepted_client_cns – list of accepted Common Names from client, if specified

• app_validator – Application folder validator.

accept_reply(client_token, reply: Message)

Accept client reply.

NOTE::

This method is to be called by the FL Engine after a client’s reply is received. Hence, it is called from the FL Engine’s message processing thread.

Parameters

• client_token – session token of the client

• reply – the reply message

client_dead(token)

Remove dead client.

Parameters

token – the session token of the client

client_heartbeat(token)

Receive client heartbeat.

Parameters

token – the session token of the client

Returns

Client.

get_client_tokens() → []

Get tokens of existing clients.

get_outgoing_requests(client_token, max_reqs=0)

Get outgoing request from a client.

NOTE::

This method is called by FL Engine to get outgoing messages to the client, so it can send them to the client.

Parameters

• client_token – session token of the client

• max_reqs – max number of requests. 0 means unlimited.

Returns

outgoing requests. A list of Message.

send_request_to_client(req: Message, client_token: str, timeout_secs=2.0) → ClientReply

send_request_to_clients(req: ~nvflare.private.admin_defs.Message, client_tokens: [<class 'str'>], timeout_secs=2.0) → [<class 'nvflare.private.fed.server.admin.ClientReply'>]

send_requests(requests: dict, timeout_secs=2.0) → [<class 'nvflare.private.fed.server.admin.ClientReply'>]

Send requests to clients.

NOTE:

This method is to be used by a Command Handler to send requests to Clients.
Hence, it is run in the Command Handler's handling thread.
This is a blocking call - returned only after all responses are received or timeout.

Parameters

• requests – A dict of requests: {client token: request or list of requests}

• timeout_secs – how long to wait for reply before timeout

Returns

A list of ClientReply

stop()

check_client_replies(replies: List[ClientReply], client_sites: List[str], command: str)

new_message(conn: Connection, topic, body) → Message