Client Protocol
最后更新于
这有帮助吗？
最后更新于
这有帮助吗？
The wire protocol used to communicate between the NATS server and clients is a simple, text-based publish/subscribe style protocol. Clients connect to and communicate with nats-server (the NATS server) through a regular TCP/IP socket using a small set of protocol operations that are terminated by a new line.
Unlike traditional messaging systems that use a binary message format that require an API to consume, the text-based NATS protocol makes it easy to implement clients in a wide variety of programming and scripting languages. In fact, refer to the topic to play with the NATS protocol for yourself using telnet.
The NATS server implements a that is fast and efficient.
Control line w/Optional Content: Each interaction between the client and server consists of a control, or protocol, line of text followed, optionally by message content. Most of the protocol messages don't require content, only PUB and MSG include payloads.
Field Delimiters: The fields of NATS protocol messages are delimited by whitespace characters '`' (space) or\t` (tab). Multiple whitespace characters will be treated as a single field delimiter.
Newlines: NATS uses CR followed by LF (CR+LF, \r\n, 0x0D0A) to terminate protocol messages. This newline sequence is also used to mark the end of the message payload in a PUB or MSG protocol message.
Subject names: Subject names, including reply subject (INBOX) names, are case-sensitive and must be non-empty alphanumeric strings with no embedded whitespace. All ascii alphanumeric characters except spaces/tabs and separators which are "." and ">" are allowed. Subject names can be optionally token-delimited using the dot character (.), e.g.:
FOO, BAR, foo.bar, foo.BAR, FOO.BAR and FOO.BAR.BAZ are all valid subject names
FOO. BAR, foo. .bar andfoo..bar are not valid subject names
A subject is comprised of 1 or more tokens. Tokens are separated by "." and can be any non space ascii alphanumeric character. The full wildcard token ">" is only valid as the last token and matches all tokens past that point. A token wildcard, "*" matches any token in the position it was listed. Wildcard tokens should only be used in a wildcard capacity and not part of a literal token.
Character Encoding: Subject names should be ascii characters for maximum interoperability. Due to language constraints and performance, some clients may support UTF-8 subject names, as may the server. No guarantees of non-ASCII support are provided.
Wildcards: NATS supports the use of wildcards in subject subscriptions.
The asterisk character (*) matches a single token at any level of the subject.
The greater than symbol (>), also known as the full wildcard, matches one or more tokens at the tail of a subject, and must be the last token. The wildcarded subject foo.> will match foo.bar or foo.bar.baz.1, but not foo.
Wildcards must be a separate token (foo.*.baz or foo.> are syntactically valid; foo*.bar, f*o.b*r and foo> are not)
For example, the wildcard subscriptions foo.*.quux and foo.> both match foo.bar.quux, but only the latter matches foo.bar.baz. With the full wildcard, it is also possible to express interest in every subject that may exist in NATS: sub > 1, limited of course by authorization settings.
The following table briefly describes the NATS protocol messages. NATS protocol operation names are case insensitive, thus SUB foo 1\r\n and sub foo 1\r\n are equivalent.
Click the name to see more detailed information, including syntax:
OP Name
Sent By
Description
Server
Sent to client after initial TCP/IP connection
Client
Sent to server to specify connection information
Client
Publish a message to a subject, with optional reply subject
Client
Subscribe to a subject (or subject wildcard)
Client
Unsubscribe (or auto-unsubscribe) from subject
Server
Delivers a message payload to a subscriber
Both
PING keep-alive message
Both
PONG keep-alive response
Server
Acknowledges well-formed protocol message in verbose mode
Server
Indicates a protocol error. May cause client disconnect.
The following sections explain each protocol message.
A client will need to start as a plain TCP connection, then when the server accepts a connection from the client, it will send information about itself, the configuration and security requirements necessary for the client to successfully authenticate with the server and exchange messages.
INFO {["option_name":option_value],...}
The valid options are as follows:
server_id: The unique identifier of the NATS server
version: The version of the NATS server
go: The version of golang the NATS server was built with
host: The IP address used to start the NATS server, by default this will be 0.0.0.0 and can be configured with -client_advertise host:port
port: The port number the NATS server is configured to listen on
max_payload: Maximum payload size, in bytes, that the server will accept from the client.
proto: An integer indicating the protocol version of the server. The server version 1.2.0 sets this to 1 to indicate that it supports the "Echo" feature.
client_id: An optional unsigned integer (64 bits) representing the internal client identifier in the server. This can be used to filter client connections in monitoring, correlate with error logs, etc...
auth_required: If this is set, then the client should try to authenticate upon connect.
tls_required: If this is set, then the client must perform the TLS/1.2 handshake. Note, this used to be ssl_required and has been updated along with the protocol from SSL to TLS.
tls_verify: If this is set, the client must provide a valid certificate during the TLS handshake.
connect_urls : An optional list of server urls that a client can connect to.
ldm: If the server supports Lame Duck Mode notifications, and the current server has transitioned to lame duck, ldm will be set to true.
The connect_urls field is a list of urls the server may send when a client first connects, and when there are changes to server cluster topology. This field is considered optional, and may be omitted based on server configuration and client protocol level.
When a NATS server cluster expands, an INFO message is sent to the client with an updated connect_urls list. This cloud-friendly feature asynchronously notifies a client of known servers, allowing it to connect to servers not originally configured.
The connect_urls will contain a list of strings with an IP and port, looking like this: "connect_urls":["10.0.0.184:4333","192.168.129.1:4333","192.168.192.1:4333"]
Below you can see a sample connection string from a telnet connection to the demo.nats.io site.
CONNECT {["option_name":option_value],...}
The valid options are as follows:
pedantic: Turns on additional strict format checking, e.g. for properly formed subjects
tls_required: Indicates whether the client requires an SSL connection.
auth_token: Client authorization token (if auth_required is set)
user: Connection username (if auth_required is set)
pass: Connection password (if auth_required is set)
name: Optional client name
lang: The implementation language of the client.
version: The version of the client.
echo: Optional boolean. If set to true, the server (version 1.2.0+) will not send originating messages from this connection to its own subscriptions. Clients should set this to true only for server supporting this feature, which is when proto in the INFO protocol is set to at least 1.
sig: In case the server has responded with a nonce on INFO, then a NATS client must use this field to reply with the signed nonce.
jwt: The JWT that identifies a user permissions and acccount.
Here is an example from the default string of the Go client:
The PUB message publishes the message payload to the given subject name, optionally supplying a reply subject. If a reply subject is supplied, it will be delivered to eligible subscribers along with the supplied payload. Note that the payload itself is optional. To omit the payload, set the payload size to 0, but the second CRLF is still required.
PUB <subject> [reply-to] <#bytes>\r\n[payload]\r\n
where:
subject: The destination subject to publish to
reply-to: The optional reply inbox subject that subscribers can use to send a response back to the publisher/requestor
#bytes: The payload size in bytes
payload: The message payload data
To publish the ASCII string message payload "Hello NATS!" to subject FOO:
PUB FOO 11\r\nHello NATS!\r\n
To publish a request message "Knock Knock" to subject FRONT.DOOR with reply subject INBOX.22:
PUB FRONT.DOOR INBOX.22 11\r\nKnock Knock\r\n
To publish an empty message to subject NOTIFY:
PUB NOTIFY 0\r\n\r\n
SUB initiates a subscription to a subject, optionally joining a distributed queue group.
SUB <subject> [queue group] <sid>\r\n
where:
subject: The subject name to subscribe to
queue group: If specified, the subscriber will join this queue group
sid: A unique alphanumeric subscription ID, generated by the client
To subscribe to the subject FOO with the connection-unique subscription identifier (sid) 1:
SUB FOO 1\r\n
To subscribe the current connection to the subject BAR as part of distribution queue group G1 with sid 44:
SUB BAR G1 44\r\n
UNSUB unsubscribes the connection from the specified subject, or auto-unsubscribes after the specified number of messages has been received.
UNSUB <sid> [max_msgs]
where:
sid: The unique alphanumeric subscription ID of the subject to unsubscribe from
max_msgs: An optional number of messages to wait for before automatically unsubscribing
The following examples concern subject FOO which has been assigned sid 1. To unsubscribe from FOO:
UNSUB 1\r\n
To auto-unsubscribe from FOO after 5 messages have been received:
UNSUB 1 5\r\n
The MSG protocol message is used to deliver an application message to the client.
MSG <subject> <sid> [reply-to] <#bytes>\r\n[payload]\r\n
where:
subject: Subject name this message was received on
sid: The unique alphanumeric subscription ID of the subject
reply-to: The inbox subject on which the publisher is listening for responses
#bytes: Size of the payload in bytes
payload: The message payload data
The following message delivers an application message from subject FOO.BAR:
MSG FOO.BAR 9 11\r\nHello World\r\n
To deliver the same message along with a reply inbox:
MSG FOO.BAR 9 INBOX.34 11\r\nHello World\r\n
PING and PONG implement a simple keep-alive mechanism between client and server. Once a client establishes a connection to the NATS server, the server will continuously send PING messages to the client at a configurable interval. If the client fails to respond with a PONG message within the configured response interval, the server will terminate its connection. If your connection stays idle for too long, it is cut off.
If the server sends a ping request, you can reply with a pong message to notify the server that you are still interested. You can also ping the server and will receive a pong reply. The ping/pong interval is configurable.
The server uses normal traffic as a ping/pong proxy, so a client that has messages flowing may not receive a ping from the server.
PING\r\n
PONG\r\n
The following example shows the demo server pinging the client and finally shutting it down.
The -ERR message is used by the server indicate a protocol, authorization, or other runtime connection error to the client. Most of these errors result in the server closing the connection.
Handling of these errors usually has to be done asynchronously.
+OK
-ERR <error message>
Some protocol errors result in the server closing the connection. Upon receiving these errors, the connection is no longer valid and the client should clean up relevant resources. These errors include:
-ERR 'Unknown Protocol Operation': Unknown protocol error
-ERR 'Attempted To Connect To Route Port': Client attempted to connect to a route port instead of the client port
-ERR 'Authorization Timeout': Client took too long to authenticate to the server after establishing a connection (default 1 second)
-ERR 'Maximum Control Line Exceeded': Message destination subject and reply subject length exceeded the maximum control line value specified by the max_control_line server option. The default is 1024 bytes.
-ERR 'Parser Error': Cannot parse the protocol message sent by the client
-ERR 'Secure Connection - TLS Required': The server requires TLS and the client does not have TLS enabled.
-ERR 'Stale Connection': The server hasn't received a message from the client, including a PONG in too long.
-ERR 'Maximum Connections Exceeded': This error is sent by the server when creating a new connection and the server has exceeded the maximum number of connections specified by the max_connections server option. The default is 64k.
-ERR 'Slow Consumer': The server pending data size for the connection has reached the maximum size (default 10MB).
Protocol error messages where the connection remains open are listed below. The client should not close the connection in these cases.
-ERR 'Invalid Subject': Client sent a malformed subject (e.g. sub foo. 90)
When using the updated client protocol (see below), INFO messages can be sent anytime by the server. This means clients with that protocol level need to be able to asynchronously handle INFO messages.
The CONNECT message is the client version of the message. Once the client has established a TCP/IP socket connection with the NATS server, and an message has been received from the server, the client may send a CONNECT message to the NATS server to provide more information about the current connection as well as security information.
verbose: Turns on protocol acknowledgements.
protocol: optional int. Sending 0 (or absent) indicates client supports original protocol. Sending 1 indicates that the client supports dynamic reconfiguration of cluster topology changes by asynchronously receiving messages with known servers it can reconnect to.
Most clients set verbose to false by default. This means that the server should not confirm each message it receives on this connection with a back to the client.
When the verbose connection option is set to true (the default value), the server acknowledges each well-formed protocol message from the client with a +OK message. Most NATS clients set the verbose option to false using the message
-ERR 'Authorization Violation': Client failed to authenticate to the server with credentials specified in the message
-ERR 'Invalid Client Protocol': Client specified an invalid protocol version in the message
-ERR 'Maximum Payload Violation': Client attempted to publish a message with a payload size that exceeds the max_payload size configured on the server. This value is supplied to the client upon connection in the initial message. The client is expected to do proper accounting of byte size to be sent to the server in order to handle this error synchronously.
-ERR 'Permissions Violation for Subscription to <subject>': The user specified in the message does not have permission to subscribe to the subject.
-ERR 'Permissions Violation for Publish to <subject>': The user specified in the message does not have permissions to publish to the subject.