ZMQ_SETSOCKOPT(3)

📦
version 4.3.5 · 11 versions
, in epel 10, fedora 43, fedora 44, opensuse-leap 16.0

ZMQ_SETSOCKOPT 💅✨

okay bestie, welcome to the 0MQ manual page for zmq_setsockopt(3) — the one where you tell a socket how to behave. buckle up, it's a long one 🎀

🌸 NAME 🌸

zmq_setsockopt — set 0MQ socket options

✨ SYNOPSIS ✨

int zmq_setsockopt (void *socket, int option_name, const void
*option_value, size_t option_len);
Caution, bestie, this one matters: literally all options — with the exception of ZMQ_SUBSCRIBE, ZMQ_UNSUBSCRIBE, ZMQ_LINGER, ZMQ_ROUTER_HANDOVER, ZMQ_ROUTER_MANDATORY, ZMQ_PROBE_ROUTER, ZMQ_XPUB_VERBOSE, ZMQ_XPUB_VERBOSER, ZMQ_REQ_CORRELATE, ZMQ_REQ_RELAXED, ZMQ_SNDHWM and ZMQ_RCVHWM — only take effect for subsequent socket bind/connects. so if you set it after you already connected… yeah, that ship has sailed 🛳️

specifically: security options take effect for subsequent bind/connect calls, and can be changed at any time to affect subsequent binds and/or connects. iykyk 💅

💖 DESCRIPTION 💖

The zmq_setsockopt() function shall set the option specified by the option_name argument to the value pointed to by the option_value argument for the 0MQ socket pointed to by the socket argument. The option_len argument is the size of the option value in bytes. For options taking a value of type "character string", the provided byte data should either contain no zero bytes, or end in a single zero byte (terminating ASCII NUL character).

and now, the moment you've all been waiting for — the options themselves 👑 The following socket options can be set with the zmq_setsockopt() function:

🧵 ZMQ_AFFINITY: Set I/O thread affinity

The ZMQ_AFFINITY option shall set the I/O thread affinity for newly created connections on the specified socket.

Affinity determines which threads from the 0MQ I/O thread pool associated with the socket's context shall handle newly created connections. A value of zero specifies no affinity, meaning that work shall be distributed fairly among all 0MQ I/O threads in the thread pool. For non-zero values, the lowest bit corresponds to thread 1, second lowest bit to thread 2 and so on. For example, a value of 3 specifies that subsequent connections on socket shall be handled exclusively by I/O threads 1 and 2.

See also zmq_init(3) for details on allocating the number of I/O threads for a specific context.

  • Option value type: uint64_t
  • Option value unit: N/A (bitmap)
  • Default value: 0
  • Applicable socket types: N/A

📚 ZMQ_BACKLOG: Set maximum length of the queue of outstanding connections

The ZMQ_BACKLOG option shall set the maximum length of the queue of outstanding peer connections for the specified socket; this only applies to connection-oriented transports. For details refer to your operating system documentation for the listen function.

  • Option value type: int
  • Option value unit: connections
  • Default value: 100
  • Applicable socket types: all, only for connection-oriented transports.

🔌 ZMQ_BINDTODEVICE: Set name of device to bind the socket to

The ZMQ_BINDTODEVICE option binds this socket to a particular device, eg. an interface or VRF. If a socket is bound to an interface, only packets received from that particular interface are processed by the socket. If device is a VRF device, then subsequent binds/connects to that socket use addresses in the VRF routing table.

Note: requires setting CAP_NET_RAW on the compiled program. no cap, you literally need the capability fr fr 🔐
  • Option value type: character string
  • Option value unit: N/A
  • Default value: not set
  • Applicable socket types: all, when using TCP or UDP transports.

🐝 ZMQ_BUSY_POLL: This removes delays caused by the interrupt and the resultant context switch.

Busy polling helps reduce latency in the network receive path by allowing socket layer code to poll the receive queue of a network device, and disabling network interrupts. This removes delays caused by the interrupt and the resultant context switch. However, it also increases CPU utilization. Busy polling also prevents the CPU from sleeping, which can incur additional power consumption.

tl;dr: faster, but your CPU is gonna be working overtime and your battery is gonna be sobbing 😭

  • Option value type: int
  • Option value unit: 0,1
  • Default value: 0
  • Applicable socket types: all

🕰️ ZMQ_CONNECT_RID: Assign the next outbound connection id

This option name is now deprecated. Use ZMQ_CONNECT_ROUTING_ID instead. ZMQ_CONNECT_RID remains as an alias for now.

deprecated alert 🚨 it's giving "we keep it around for the vibes but please move on"

🎯 ZMQ_CONNECT_ROUTING_ID: Assign the next outbound routing id

The ZMQ_CONNECT_ROUTING_ID option sets the peer id of the peer connected via the next zmq_connect() call, such that that connection is immediately ready for data transfer with the given routing id. This option applies only to the first subsequent call to zmq_connect(), zmq_connect() calls thereafter use the default connection behaviour.

Typical use is to set this socket option ahead of each zmq_connect() call. Each connection MUST be assigned a unique routing id. Assigning a routing id that is already in use is not allowed.

Useful when connecting ROUTER to ROUTER, or STREAM to STREAM, as it allows for immediate sending to peers. Outbound routing id framing requirements for ROUTER and STREAM sockets apply.

The routing id must be from 1 to 255 bytes long and MAY NOT start with a zero byte (such routing ids are reserved for internal use by the 0MQ infrastructure).

  • Option value type: binary data
  • Option value unit: N/A
  • Default value: NULL
  • Applicable socket types: ZMQ_ROUTER, ZMQ_STREAM

🫧 ZMQ_CONFLATE: Keep only last message

If set, a socket shall keep only one message in its inbound/outbound queue, this message being the last message received/the last message to be sent. Ignores ZMQ_RCVHWM and ZMQ_SNDHWM options. Does not support multi-part messages, in particular, only one part of it is kept in the socket internal queue.

Note: if recv is not called on the inbound socket, the queue and memory will grow with each message received. Use zmq_getsockopt(3) with ZMQ_EVENTS to trigger the conflation of the messages. memory leak vibes, be careful bestie 💀
  • Option value type: int
  • Option value unit: boolean
  • Default value: 0 (false)
  • Applicable socket types: ZMQ_PULL, ZMQ_PUSH, ZMQ_SUB, ZMQ_PUB, ZMQ_DEALER

⏳ ZMQ_CONNECT_TIMEOUT: Set connect() timeout

Sets how long to wait before timing-out a connect() system call. The connect() system call normally takes a long time before it returns a time out error. Setting this option allows the library to time out the call at an earlier interval.

no more waiting around for a connect that's never gonna text you back 🙈

  • Option value type: int
  • Option value unit: milliseconds
  • Default value: 0 (disabled)
  • Applicable socket types: all, when using TCP transports.

🔑 ZMQ_CURVE_PUBLICKEY: Set CURVE public key

Sets the socket's long term public key. You must set this on CURVE client sockets, see zmq_curve(7). You can provide the key as 32 binary bytes, or as a 40-character string encoded in the Z85 encoding format and terminated in a null byte. The public key must always be used with the matching secret key. To generate a public/secret key pair, use zmq_curve_keypair(3). To derive the public key from a secret key, use zmq_curve_public(3).

Note: an option value size of 40 is supported for backwards compatibility, though is deprecated.
  • Option value type: binary data or Z85 text string
  • Option value size: 32 or 41
  • Default value: NULL
  • Applicable socket types: all, when using TCP transport

🤫 ZMQ_CURVE_SECRETKEY: Set CURVE secret key

Sets the socket's long term secret key. You must set this on both CURVE client and server sockets, see zmq_curve(7). You can provide the key as 32 binary bytes, or as a 40-character string encoded in the Z85 encoding format and terminated in a null byte. To generate a public/secret key pair, use zmq_curve_keypair(3). To derive the public key from a secret key, use zmq_curve_public(3).

Note: an option value size of 40 is supported for backwards compatibility, though is deprecated.
  • Option value type: binary data or Z85 text string
  • Option value size: 32 or 41
  • Default value: NULL
  • Applicable socket types: all, when using TCP transport

👑 ZMQ_CURVE_SERVER: Set CURVE server role

Defines whether the socket will act as server for CURVE security, see zmq_curve(7). A value of 1 means the socket will act as CURVE server. A value of 0 means the socket will not act as CURVE server, and its security role then depends on other option settings. Setting this to 0 shall reset the socket security to NULL. When you set this you must also set the server's secret key using the ZMQ_CURVE_SECRETKEY option. A server socket does not need to know its own public key.

  • Option value type: int
  • Option value unit: 0, 1
  • Default value: 0
  • Applicable socket types: all, when using TCP transport

🗝️ ZMQ_CURVE_SERVERKEY: Set CURVE server key

Sets the socket's long term server key. You must set this on CURVE client sockets, see zmq_curve(7). You can provide the key as 32 binary bytes, or as a 40-character string encoded in the Z85 encoding format and terminated in a null byte. This key must have been generated together with the server's secret key. To generate a public/secret key pair, use zmq_curve_keypair(3).

Note: an option value size of 40 is supported for backwards compatibility, though is deprecated.
  • Option value type: binary data or Z85 text string
  • Option value size: 32 or 41
  • Default value: NULL
  • Applicable socket types: all, when using TCP transport

👋 ZMQ_DISCONNECT_MSG: set a disconnect message that the socket will generate when accepted peer disconnect

When set, the socket will generate a disconnect message when accepted peer has been disconnected. You may set this on ROUTER, SERVER and PEER sockets. The combination with ZMQ_HEARTBEAT_IVL is powerful and simplify protocols, when heartbeat recognize a connection drop it will generate a disconnect message that can match the protocol of the application.

  • Option value type: binary data
  • Option value unit: N/A
  • Default value: NULL
  • Applicable socket types: ZMQ_ROUTER, ZMQ_SERVER and ZMQ_PEER

🥺 ZMQ_HICCUP_MSG: set a hiccup message that the socket will generate when connected peer temporarily disconnect

When set, the socket will generate a hiccup message when connect peer has been disconnected. You may set this on DEALER, CLIENT and PEER sockets. The combination with ZMQ_HEARTBEAT_IVL is powerful and simplify protocols, when heartbeat recognize a connection drop it will generate a hiccup message that can match the protocol of the application.

…and finishing off the option from before 💅 its deets:

  • Option value type — binary data
  • Option value unit — N/A
  • Default value — NULL
  • Applicable socket types — ZMQ_DEALER, ZMQ_CLIENT and ZMQ_PEER

🔐 ZMQ_GSSAPI_PLAINTEXT — Disable GSSAPI encryption

Bestie, this one decides whether the chatter on your socket gets encrypted or not — peep zmq_gssapi(7) for the whole lore. Set it to 1 and the comms are plaintext (aka out in the open, no cap). Set it to 0 and the comms are encrypted. That's it, that's the whole vibe.

  • Option value type — int
  • Option value unit — 0, 1
  • Default value — 0 (false)
  • Applicable socket types — all, when using TCP transport

👑 ZMQ_GSSAPI_PRINCIPAL — Set name of GSSAPI principal

This sets the name of the principal for whom GSSAPI credentials should be acquired. Like, whose ✨identity✨ are we getting the keys for.

  • Option value type — character string
  • Option value unit — N/A
  • Default value — not set
  • Applicable socket types — all, when using TCP transport

🎀 ZMQ_GSSAPI_SERVER — Set GSSAPI server role

Does your socket wanna be the GSSAPI server or the GSSAPI client? See zmq_gssapi(7) for the full story. Value 1 means server energy, value 0 means client energy. Pick your main character role.

  • Option value type — int
  • Option value unit — 0, 1
  • Default value — 0 (false)
  • Applicable socket types — all, when using TCP transport

🌸 ZMQ_GSSAPI_SERVICE_PRINCIPAL — Set name of GSSAPI service principal

Sets the name of the principal of the GSSAPI server that a GSSAPI client is tryna connect to. Like, whose door are we knocking on.

  • Option value type — character string
  • Option value unit — N/A
  • Default value — not set
  • Applicable socket types — all, when using TCP transport

🦋 ZMQ_GSSAPI_SERVICE_PRINCIPAL_NAMETYPE — Set name type of service principal

Okay so this sets the name type of the GSSAPI service principal. Here's the menu:

  • ZMQ_GSSAPI_NT_HOSTBASED (0) — the name you gave with ZMQ_GSSAPI_SERVICE_PRINCIPAL is interpreted as a host based name.
  • ZMQ_GSSAPI_NT_USER_NAME (1) — it's interpreted as a local user name.
  • ZMQ_GSSAPI_NT_KRB5_PRINCIPAL (2) — it's interpreted as an unparsed principal name string (valid only with the krb5 GSSAPI mechanism, iykyk).
  • Option value type — int
  • Option value unit — 0, 1, 2
  • Default value — 0 (ZMQ_GSSAPI_NT_HOSTBASED)
  • Applicable socket types — all, when using TCP or IPC transport

💜 ZMQ_GSSAPI_PRINCIPAL_NAMETYPE — Set name type of principal

Same energy as above but for the GSSAPI principal itself, no service in the name. Sets the name type of the GSSAPI principal:

  • ZMQ_GSSAPI_NT_HOSTBASED (0) — the name specified with ZMQ_GSSAPI_PRINCIPAL is interpreted as a host based name.
  • ZMQ_GSSAPI_NT_USER_NAME (1) — interpreted as a local user name.
  • ZMQ_GSSAPI_NT_KRB5_PRINCIPAL (2) — interpreted as an unparsed principal name string (valid only with the krb5 GSSAPI mechanism).
  • Option value type — int
  • Option value unit — 0, 1, 2
  • Default value — 0 (ZMQ_GSSAPI_NT_HOSTBASED)
  • Applicable socket types — all, when using TCP or IPC transport

🤝 ZMQ_HANDSHAKE_IVL — Set maximum handshake interval

So the ZMQ_HANDSHAKE_IVL option sets the maximum handshake interval for the socket you point it at. And what even is a handshake? It's the exchange of socket configuration info — socket type, routing id, security — that happens the very first time a connection is opened, and it's only a thing for connection-oriented transports. If the handshake doesn't wrap up inside the time you configured, the connection gets closed, periodt. A value of 0 means no handshake time limit at all — free spirit behavior.

  • Option value type — int
  • Option value unit — milliseconds
  • Default value — 30000
  • Applicable socket types — all but ZMQ_STREAM, only for connection-oriented transports

🫶 ZMQ_HELLO_MSG — set an hello message that will be sent when a new peer connect

Set this and your socket will automatically send a hello message the moment a new connection is made or accepted. You may set this on DEALER, ROUTER, CLIENT, SERVER and PEER sockets. And honestly? The combo with ZMQ_HEARTBEAT_IVL is powerful 🔥 — it simplifies protocols so much, because now the heartbeat and the hello message can be left out of your protocols entirely and zeromq handles them for you. Obsessed.

  • Option value type — binary data
  • Option value unit — N/A
  • Default value — NULL
  • Applicable socket types — ZMQ_ROUTER, ZMQ_DEALER, ZMQ_CLIENT, ZMQ_SERVER and ZMQ_PEER

💓 ZMQ_HEARTBEAT_IVL — Set interval between sending ZMTP heartbeats

The ZMQ_HEARTBEAT_IVL option sets the interval between sending ZMTP heartbeats for the specified socket. If it's set and it's greater than 0, then a PING ZMTP command gets sent every ZMQ_HEARTBEAT_IVL milliseconds. Cute, consistent, reliable. 💖

  • Option value type — int
  • Option value unit — milliseconds
  • Default value — 0
  • Applicable socket types — all, when using connection-oriented transports

⏳ ZMQ_HEARTBEAT_TIMEOUT — Set timeout for ZMTP heartbeats

The ZMQ_HEARTBEAT_TIMEOUT option sets how long to wait before timing out a connection after you've sent a PING ZMTP command and gotten no traffic back. Heads up: this option is only valid if ZMQ_HEARTBEAT_IVL is also set, and is greater than 0. The connection will time out if there's no traffic received after sending the PING command — but plot twist, the received traffic does not have to be a PONG command. Any received traffic at all cancels the timeout. Slay.

  • Option value type — int
  • Option value unit — milliseconds
  • Default value — 0 normally, ZMQ_HEARTBEAT_IVL if it is set
  • Applicable socket types — all, when using connection-oriented transports

⌛ ZMQ_HEARTBEAT_TTL — Set the TTL value for ZMTP heartbeats

The ZMQ_HEARTBEAT_TTL option sets the timeout on the remote peer for ZMTP heartbeats. If this option is greater than 0, the remote side will time out the connection if it doesn't receive any more traffic within the TTL period. Ngl, this option does nothing at all if ZMQ_HEARTBEAT_IVL isn't set or is 0. And internally, this value gets rounded down to the nearest decisecond — so any value less than 100 will have no effect. 💀

  • Option value type — int
  • Option value unit — milliseconds
  • Default value — 0
  • Maximum value — 6553599 (which is 2^16-1 deciseconds)
  • Applicable socket types — all, when using connection-oriented transports

🪪 ZMQ_IDENTITY — Set socket identity

Okay so this option name is now deprecated, bestie. Use ZMQ_ROUTING_ID instead. ZMQ_IDENTITY is sticking around as an alias for now.

🚦 ZMQ_IMMEDIATE — Queue messages only to completed connections

By default, queues will fill on outgoing connections even if the connection hasn't completed yet. And that can lead to "lost" messages on sockets with round-robin routing (REQ, PUSH, DEALER). Not the vibe. If you set this option to 1, messages will be queued only to completed connections. Yes, this will cause the socket to block if there are no other connections — but it prevents queues from filling on pipes that are still waiting to connect. Trade-off, but a good one.

  • Option value type — int
  • Option value unit — boolean
  • Default value — 0 (false)
  • Applicable socket types — all, only for connection-oriented transports.

🙃 ZMQ_INVERT_MATCHING — Invert message filtering

This one flips the filtering behavior of PUB-SUB sockets when set to 1. Iconic chaos.

On PUB and XPUB sockets, it causes messages to be sent to all connected sockets except those subscribed to a prefix that matches the message. On SUB sockets, it causes only incoming messages that do not match any of the socket's subscriptions to be received by the user.

⚠️ Whenever ZMQ_INVERT_MATCHING is set to 1 on a PUB socket, all SUB sockets connecting to it must also have the option set to 1. If you don't, the SUB sockets will reject everything the PUB socket sends them. (XSUB sockets don't need to do this, because they don't filter incoming messages — they're built different.)
  • Option value type — int
  • Option value unit — 0,1
  • Default value — 0
  • Applicable socket types — ZMQ_PUB, ZMQ_XPUB, ZMQ_SUB

🌈 ZMQ_IPV6 — Enable IPv6 on socket

Sets the IPv6 option for the socket. A value of 1 means IPv6 is enabled on the socket, while 0 means the socket will use only IPv4. And here's the cute part: when IPv6 is enabled, the socket will connect to — or accept connections from — both IPv4 and IPv6 hosts. Both. Multitasking queen. 👑

  • Option value type — int
  • Option value unit — boolean
  • Default value — 0 (false)
  • Applicable socket types — all, when using TCP transports.

🕰️ ZMQ_LINGER — Set linger period for socket shutdown

The ZMQ_LINGER option sets the linger period for the specified socket. So what does the linger period even do? It determines how long pending messages that haven't been sent to a peer yet will linger in memory after the socket is disconnected with zmq_disconnect(3) or closed with zmq_close(3) — and it also affects the termination of the socket's context with zmq_ctx_term(3). Here's the breakdown of the different behaviours:

  • A value of -1 specifies an infinite linger period. Pending messages shall not be discarded after a call to zmq_disconnect() or zmq_close(); attempting to terminate the socket's context with zmq_ctx_term() shall block until all pending messages have been sent to a peer.
  • The value of 0 specifies no linger period. Pending messages shall be discarded immediately after a call to zmq_disconnect() or zmq_close().
  • Positive values specify an upper bound for the linger period in milliseconds. Pending messages shall not be discarded after a call to zmq_disconnect() or zmq_close(); attempting to terminate the socket's context with zmq_ctx_term() shall block until either all pending messages have been sent to a peer, or the linger period expires, after which any pending messages shall be discarded.
  • Option value type — int
  • Option value unit — milliseconds
  • Default value — -1 (infinite)
  • Applicable socket types — all

📏 ZMQ_MAXMSGSIZE — Maximum acceptable inbound message size

Limits the size of the inbound message. If a peer sends a message larger than ZMQ_MAXMSGSIZE, it gets disconnected — no negotiation, no discussion, just gone. 💀 A value of -1 means no limit at all.

  • Option value type — int64_t
  • Option value unit — bytes
  • Default value — -1
  • Applicable socket types — all

✨ ZMQ_METADATA: Add application metadata properties to a socket ✨

ok so ZMQ_METADATA is the one that lets you attach your own lil metadata properties to a socket, and those get swapped with your peers during connection setup 💫. Here's how you write one: it's a string, split up by a colon, and it starts with the metadata property, then the metadata value — like literally "X-key:value", that's the shape 🎀.

The property names are capped at 255 characters max and they MUST be prefixed with "X-" — no exceptions bestie. And if you want multiple application metadata properties on one socket? Just call zmq_setsockopt() over and over, once per property. Obsessed with the flexibility ngl 💅.

One thing tho — the argument is a null-terminated string, so if you've got binary data you gotta encode it first before adding it, e.g. with Z85 (zmq_z85_encode(3)) 🦋.

Note: in DRAFT state, not yet available in stable releases. 🙈
  • Option value type — character string
  • Option value unit — N/A
  • Default value — not set
  • Applicable socket types — all

🌈 ZMQ_MULTICAST_HOPS: Maximum network hops for multicast packets 🌈

This one sets the time-to-live field in every single multicast packet sent from this socket. Default is 1, which means your multicast packets literally do not leave the local network — they're homebodies fr fr 🏠.

  • Option value type — int
  • Option value unit — network hops
  • Default value — 1
  • Applicable socket types — all, when using multicast transports

🎀 ZMQ_MULTICAST_MAXTPDU: Maximum transport data unit size for multicast packets 🎀

Sets the max transport data unit size used for outbound multicast packets ✨.

Careful bestie: this must be set at or below the minimum Maximum Transmission Unit (MTU) for all network paths over which multicast reception is required. No going over, okay? 💅
  • Option value type — int
  • Option value unit — bytes
  • Default value — 1500
  • Applicable socket types — all, when using multicast transports

🔐 ZMQ_PLAIN_PASSWORD: Set PLAIN security password 🔐

Sets the password for outgoing connections over TCP or IPC. Set it to a non-null value and the security mechanism used for connections shall be PLAIN, see zmq_plain(7). Set it to a null value and the security mechanism used for connections shall be NULL, see zmq_null(3). The duality of it all 😭

  • Option value type — character string
  • Option value unit — N/A
  • Default value — not set
  • Applicable socket types — all, when using TCP transport

👑 ZMQ_PLAIN_SERVER: Set PLAIN server role 👑

Defines whether the socket is gonna act as server for PLAIN security, see zmq_plain(7). A value of 1 means yes, this socket is the PLAIN server, main character energy 👑. A value of 0 means nope, it is not the PLAIN server, and its security role then depends on other option settings. And setting this to 0 shall reset the socket security to NULL — full glow-down 💀.

  • Option value type — int
  • Option value unit — 0, 1
  • Default value — 0
  • Applicable socket types — all, when using TCP transport

🌸 ZMQ_PLAIN_USERNAME: Set PLAIN security username 🌸

Sets the username for outgoing connections over TCP or IPC. Non-null value → the security mechanism used for connections shall be PLAIN, see zmq_plain(7). Null value → the security mechanism used for connections shall be NULL, see zmq_null(3). Same energy as the password one, they're a matching set 🫶

  • Option value type — character string
  • Option value unit — N/A
  • Default value — not set
  • Applicable socket types — all, when using TCP transport

🧚 ZMQ_USE_FD: Set the pre-allocated socket file descriptor 🧚

Set this to a positive integer value before zmq_bind is called on the socket, and the socket shall use that corresponding file descriptor for connections over TCP or IPC instead of allocating a new one. Super useful for writing systemd socket activated services btw ✨. Set it to -1 (the default) and a new file descriptor will be allocated instead — which is, yep, the default behaviour.

Big note, read this twice bestie: if set after calling zmq_bind, this option shall have no effect. NOTE: the file descriptor passed through MUST have been ran through the "bind" and "listen" system calls beforehand. Also, socket option that would normally be passed through zmq_setsockopt like TCP buffers length, IP_TOS or SO_REUSEADDR MUST be set beforehand by the caller, as they must be set before the socket is bound. 💀 no skips.
  • Option value type — int
  • Option value unit — file descriptor
  • Default value — -1
  • Applicable socket types — all bound sockets, when using IPC or TCP transport

💜 ZMQ_PRIORITY: Set the Priority on socket 💜

Sets the protocol-defined priority for all packets to be sent on this socket — where supported by the OS, so vibe check your kernel first 🙈. On Linux, values greater than 6 require admin capability (CAP_NET_ADMIN). So like, you need the receipts 💅

  • Option value type — int
  • Option value unit — >0
  • Default value — 0
  • Applicable socket types — all, only for connection-oriented transports

🌟 ZMQ_PROBE_ROUTER: bootstrap connections to ROUTER sockets 🌟

Set this to 1 and the socket will automatically send an empty message whenever a new connection is made or accepted 👀. You can set it on REQ, DEALER, or ROUTER sockets connected to a ROUTER socket. The application must filter out such empty messages — that's on you, not on 0MQ. What ZMQ_PROBE_ROUTER in effect does is give the ROUTER application an event signaling the arrival of a new peer 🎉 iconic honestly.

Do NOT set this option on a socket that talks to any other socket types: the results are undefined. Not "a little weird", not "probably fine" — undefined. Periodt.
  • Option value type — int
  • Option value unit — 0, 1
  • Default value — 0
  • Applicable socket types — ZMQ_ROUTER, ZMQ_DEALER, ZMQ_REQ

💫 ZMQ_RATE: Set multicast data rate 💫

ZMQ_RATE shall set the maximum send or receive data rate for multicast transports such as zmq_pgm(7) using the specified socket. Rate limiting but make it multicast 💅

  • Option value type — int
  • Option value unit — kilobits per second
  • Default value — 100
  • Applicable socket types — all, when using multicast transports

🍓 ZMQ_RCVBUF: Set kernel receive buffer size 🍓

ZMQ_RCVBUF shall set the underlying kernel receive buffer size for the socket to the specified size in bytes 🧺. A value of -1 means leave the OS default unchanged — not my business, not my problem. For the nitty gritty details refer to your operating system documentation for the SO_RCVBUF socket option 📚

  • Option value type — int
  • Option value unit — bytes
  • Default value — -1
  • Applicable socket types — all

🦋 ZMQ_RCVHWM: Set high water mark for inbound messages 🦋

ZMQ_RCVHWM shall set the high water mark for inbound messages on the specified socket. And what is a high water mark? It's a hard limit on the maximum number of outstanding messages 0MQ shall queue in memory for any single peer that the specified socket is communicating with 🌊. A value of zero means no limit — yolo mode, no cap.

If this limit has been reached, the socket shall enter an exceptional state and depending on the socket type, 0MQ shall take appropriate action such as blocking or dropping sent messages 😭. Refer to the individual socket descriptions in zmq_socket(3) for details on the exact action taken for each socket type.

Note: 0MQ does not guarantee that the socket will be able to queue as many as ZMQ_RCVHWM messages, and the actual limit may be lower or higher, depending on socket transport. A notable example is for sockets using TCP transport; see zmq_tcp(7). So don't come for 0MQ when the number's off, it said what it said 🙈
  • Option value type — int
  • Option value unit — messages
  • Default value — 1000
  • Applicable socket types — all

⏰ ZMQ_RCVTIMEO: Maximum time before a recv operation returns with EAGAIN ⏰

Sets the timeout for receive operation on the socket. Here's the rundown:

  • Value 0 → zmq_recv(3) will return immediately, with an EAGAIN error if there's no message to receive. Snappy 💨
  • Value -1 → it will block until a message is available. Patient queen 🧘‍♀️
  • Any other value → it will wait for a message for that amount of time before returning with an EAGAIN error.
  • Option value type — int
  • Option value unit — milliseconds
  • Default value — -1 (infinite)
  • Applicable socket types — all

🔁 ZMQ_RECONNECT_IVL: Set reconnection interval 🔁

ZMQ_RECONNECT_IVL shall set the initial reconnection interval for the specified socket. The reconnection interval is the period 0MQ shall wait between attempts to reconnect disconnected peers when using connection-oriented transports 🫂. The value -1 means no reconnection — it's giving "we're done here" 💀

Note: the reconnection interval may be randomized by 0MQ to prevent reconnection storms in topologies with a large number of peers per socket. Anti-storm technology, we love to see it 🌈
  • Option value type — int
  • Option value unit — milliseconds
  • Default value — 100
  • Applicable socket types — all, only for connection-oriented transports

📈 ZMQ_RECONNECT_IVL_MAX: Set max reconnection interval 📈

ZMQ_RECONNECT_IVL_MAX shall set the max reconnection interval for the specified socket. 0MQ shall wait at most the configured interval between reconnection attempts. The interval grows exponentionally (i.e.: it is doubled) with each attempt until it reaches ZMQ_RECONNECT_IVL_MAX — yes the manpage really does spell it "exponentionally", we're keeping it, character for character bestie 💅. The default value means that the reconnect interval is based exclusively on ZMQ_RECONNECT_IVL and no exponential backoff is performed. Slow and steady, no escalation 🧚

Note: value has to be greater or equal than ZMQ_RECONNECT_IVL, or else it will be ignored. Read that again 🫶
  • Option value type — int
  • Option value unit — milliseconds
  • Default value — 0 (ZMQ_RECONNECT_IVL will be used)
  • Applicable socket types — all, only for connection-oriented transports

🛑 ZMQ_RECONNECT_STOP: Set condition where reconnection will stop 🛑

ZMQ_RECONNECT_STOP shall set the conditions under which automatic reconnection will stop. This can be useful when a process binds to a wild-card port, where the OS supplies an ephemeral port 🌸.

The ZMQ_RECONNECT_STOP_CONN_REFUSED option will stop reconnection when 0MQ receives the ECONNREFUSED return code from the connect. This indicates that there is no code bound to the specified endpoint — nobody's home, so we stop knocking 🚪

The ZMQ_RECONNECT_STOP_HANDSHAKE_FAILED option will stop reconnection if the 0MQ handshake fails. This can be used to detect and/or prevent errant connection attempts to non-0MQ sockets 👀. Note that when specifying this option you may also want to set ZMQ_HANDSHAKE_IVL — the default handshake interval is 30000 (30 seconds), which is typically too large. That's a whole 30 seconds of your life, bestie, set it lower 💀

ok so the whole point of ZMQ_RECONNECT_STOP_AFTER_DISCONNECT is: the second somebody calls zmq_disconnect(), the reconnecting era is over, bestie. No more reconnect attempts. Periodt. And that's genuinely a slay when the user's own request failed (classic server-not-ready moment) — because why would the socket keep trying to reconnect when the user actively dipped? It doesn't need to, babes. Let it rest. 💅

Note: this one is in DRAFT state, so it is not yet available in stable releases. Do not build your whole personality around it just yet.
  • Option value type — int
  • Option value unit — 0, ZMQ_RECONNECT_STOP_CONN_REFUSED, ZMQ_RECONNECT_STOP_HANDSHAKE_FAILED, ZMQ_RECONNECT_STOP_CONN_REFUSED | ZMQ_RECONNECT_STOP_HANDSHAKE_FAILED
  • Default value — 0
  • Applicable socket types — all, only for connection-oriented transports (ZMQ_HANDSHAKE_IVL is not applicable for ZMQ_STREAM sockets)

🌈 ZMQ_RECOVERY_IVL: Set multicast recovery interval 🧚

ZMQ_RECOVERY_IVL sets the recovery interval for multicast transports on the socket you're holding, and honestly it's the "how long can you ghost before it's unfixable" setting. The recovery interval = the maximum time in milliseconds that a receiver can be absent from a multicast group before unrecoverable data loss happens. Like, after that window, the data is gone gone. 💀

Caution: be SO careful with big recovery intervals, bestie — the data needed for recovery gets held in memory. The original's own example, and I am not exaggerating: a 1 minute recovery interval at a data rate of 1Gbps requires a 7GB in-memory buffer. Seven. Gigabytes. Your RAM could never. 😭
  • Option value type — int
  • Option value unit — milliseconds
  • Default value — 10000
  • Applicable socket types — all, when using multicast transports

✨ ZMQ_REQ_CORRELATE: match replies with requests 🎀

So by default, REQ sockets are just vibing off ordering to match up requests and responses, and honestly? Usually that's enough, no notes. But when you set this to 1, the REQ socket gets a whole extra frame at the front of outgoing messages containing a request id — so the full message becomes (request id, 0, user frames...). And on the way in, the REQ socket discards every incoming message that doesn't begin with those two frames. Gatekeeping, but make it protocol. 🚪

  • Option value type — int
  • Option value unit — 0, 1
  • Default value — 0
  • Applicable socket types — ZMQ_REQ

💖 ZMQ_REQ_RELAXED: relax strict alternation between request and reply 💅

Default REQ behaviour is strict alternation, iykyk: you cannot start a new request with zmq_send(3) until the reply to the previous one has come home. Set this to 1 and suddenly you can send another message, and any previous replies get discarded if there are any. The request-reply state machine resets, and a brand new request goes out to the next available peer. Main character energy, honestly. ✨

Warning: if you set this to 1, also turn on ZMQ_REQ_CORRELATE so requests and replies match up correctly. Otherwise a late reply to an aborted request can get reported as the reply to the superseding request — which is a whole mess. 😭
  • Option value type — int
  • Option value unit — 0, 1
  • Default value — 0
  • Applicable socket types — ZMQ_REQ

🦋 ZMQ_ROUTER_HANDOVER: handle duplicate client routing ids on ROUTER sockets 👑

Two clients, same routing id, both connecting to a ROUTER — what happens next is entirely up to the ZMQ_ROUTER_HANDOVER setting, bestie. If it's not set (or set to the default of zero), the ROUTER socket shall reject clients trying to connect with an already-used routing id. Rude but fair. If it is set to 1, the ROUTER socket shall hand-over the connection to the new client and disconnect the existing one. Out with the old, in with the new. 💅

  • Option value type — int
  • Option value unit — 0, 1
  • Default value — 0
  • Applicable socket types — ZMQ_ROUTER

🔥 ZMQ_ROUTER_MANDATORY: accept only routable messages on ROUTER sockets ⭐

This one controls what a ROUTER socket does when it meets an unroutable message. A value of 0 is the default and it silently discards the message when it can't be routed or when the peer's SNDHWM is reached — just, gone, no drama, no notification. A value of 1 returns an EHOSTUNREACH error code if the message can't be routed, or EAGAIN if the SNDHWM is reached and ZMQ_DONTWAIT was used. Without ZMQ_DONTWAIT it'll block until SNDTIMEO is reached or a spot opens up in the send queue. 🫶

When ZMQ_ROUTER_MANDATORY is set to 1, ZMQ_POLLOUT events get generated if one or more messages can be sent to at least one of the peers. When it's set to 0, the socket generates a ZMQ_POLLOUT event on every call to zmq_poll resp. zmq_poller_wait_all. Yeah. Every single one. 💀

  • Option value type — int
  • Option value unit — 0, 1
  • Default value — 0
  • Applicable socket types — ZMQ_ROUTER

🌸 ZMQ_ROUTER_RAW: switch ROUTER socket to raw mode 🧚

Sets raw mode on the ROUTER when set to 1. When the ROUTER socket is in raw mode, and when using the tcp:// transport, it will read and write TCP data without 0MQ framing — which is how your 0MQ applications get to talk to non-0MQ applications. Iconic crossover episode. 🌈

While in raw mode: you cannot set explicit identities, and the ZMQ_SNDMORE flag is ignored when sending data messages. Also, in raw mode you can close a specific connection by sending it a zero-length message (following the routing id frame). 🎀

Note: this option is deprecated, bestie. Please use ZMQ_STREAM sockets instead. It's giving "we don't do that anymore." 💅
  • Option value type — int
  • Option value unit — 0, 1
  • Default value — 0
  • Applicable socket types — ZMQ_ROUTER

💜 ZMQ_ROUTING_ID: Set socket routing id 🎀

ZMQ_ROUTING_ID sets the routing id of the specified socket when connecting to a ROUTER socket.

A routing id must be at least one byte and at most 255 bytes long — no exceptions, that's the range, fr fr. And identities starting with a zero byte are reserved for use by the 0MQ infrastructure, so hands off those, they're not yours. 🙈

If two clients use the same routing id when connecting to a ROUTER, the results depend on the ZMQ_ROUTER_HANDOVER option setting. If that's not set (or set to the default of zero), the ROUTER socket shall reject clients trying to connect with an already-used routing id. If that option is set to 1, the ROUTER socket shall hand-over the connection to the new client and disconnect the existing one.

  • Option value type — binary data
  • Option value unit — N/A
  • Default value — NULL
  • Applicable socket types — ZMQ_REQ, ZMQ_REP, ZMQ_ROUTER, ZMQ_DEALER.

🌈 ZMQ_SNDBUF: Set kernel transmit buffer size 💖

ZMQ_SNDBUF sets the underlying kernel transmit buffer size for the socket to the specified size in bytes. A value of -1 means leave the OS default unchanged — untouched, unbothered. For the details, please go read your operating system documentation for the SO_SNDBUF socket option, because that's where the real lore lives. 🦋

  • Option value type — int
  • Option value unit — bytes
  • Default value — -1
  • Applicable socket types — all

⭐ ZMQ_SNDHWM: Set high water mark for outbound messages 👑

ZMQ_SNDHWM sets the high water mark for outbound messages on the socket. And what is a high water mark, you ask? It's a hard limit on the maximum number of outstanding messages 0MQ shall queue in memory for any single peer that the specified socket is communicating with. A value of zero means no limit — unlimited, unhinged, no cap. 💅

If this limit has been reached, the socket shall enter an exceptional state and, depending on the socket type, 0MQ shall take appropriate action such as blocking or dropping sent messages. Go check the individual socket descriptions in zmq_socket(3) for the exact action taken for each socket type — the vibes differ per socket, bestie.

Note: 0MQ does not guarantee that the socket will accept as many as ZMQ_SNDHWM messages, and the actual limit may be as much as 90% lower depending on the flow of messages on the socket. The socket may even be able to accept more messages than the ZMQ_SNDHWM threshold — a notable example being sockets using TCP transport; see zmq_tcp(7). So like… it's a suggestion that's also a hard limit. Make it make sense. 😭
  • Option value type — int
  • Option value unit — messages
  • Default value — 1000
  • Applicable socket types — all

🎀 ZMQ_SNDTIMEO: Maximum time before a send operation returns with EAGAIN 🫶

Sets the timeout for send operations on the socket, and the three modes are very much a personality test:

  • If the value is 0, zmq_send(3) will return immediately, with an EAGAIN error if the message cannot be sent. Instant. No waiting. Not even a little.
  • If the value is -1, it will block until the message is sent. Devotion. Loyalty. 🥺
  • For all other values, it will try to send the message for that amount of time before returning with an EAGAIN error.
  • Option value type — int
  • Option value unit — milliseconds
  • Default value — -1 (infinite)
  • Applicable socket types — all

🦋 ZMQ_SOCKS_PROXY: Set SOCKS5 proxy address 🌸

Sets the SOCKS5 proxy address that shall be used by the socket for the TCP connection(s). Supported authentication methods are: no authentication, or basic authentication when you set it up with ZMQ_SOCKS_USERNAME.

And here's the cute part: if the endpoints are domain names instead of addresses, they shall not be resolved — they get forwarded unchanged to the SOCKS proxy service in the client connection request message (address type 0x03 domain name). Let the proxy do the work, bestie. 💅

  • Option value type — character string
  • Option value unit — N/A
  • Default value — not set
  • Applicable socket types — all, when using TCP transport

💜 ZMQ_SOCKS_USERNAME: Set SOCKS username and select basic authentication 🔥

Sets the username for an authenticated connection to the SOCKS5 proxy. If you set this to a non-null and non-empty value, the authentication method used for the SOCKS5 connection shall be basic authentication — and in that case, go use the ZMQ_SOCKS_PASSWORD option to set the password. 🔑

If you set this to a null value or an empty value, the authentication method shall be no authentication, which is the default. So really, this option is also the "which auth method am I doing" switch. Two jobs, one flag, we love a multitasker. ✨

  • Option value type — character string
  • Option value unit — N/A
  • Default value — not set
  • Applicable socket types — all, when using TCP transport

🍓 ZMQ_SOCKS_PASSWORD: Set SOCKS basic authentication password 🔐

Sets the password for authenticating to the SOCKS5 proxy server. This is used only when the SOCKS5 authentication method has been set to basic authentication through the ZMQ_SOCKS_USERNAME option — so no username, no password usage, them's the rules. Setting this to a null value (the default) is equivalent to an empty password string. 🎀

  • Option value type — character string
  • Option value unit — N/A
  • Default value — not set
  • Applicable socket types — all, when using TCP transport

💖 ZMQ_STREAM_NOTIFY: send connect and disconnect notifications 🔔

Enables connect and disconnect notifications on a STREAM socket, when set to 1. When notifications are enabled, the socket delivers a zero-length message when a peer connects or disconnects. It's like a little "hey, they're here!" and "oop, they left" ping, and honestly it's the politest thing a socket can do. 🫶

  • Option value type — int
  • Option value unit — 0, 1
  • Default value — 1
  • Applicable socket types — ZMQ_STREAM

🌸 ZMQ_SUBSCRIBE: Establish message filter

Okay bestie, listen 💖 — ZMQ_SUBSCRIBE is how you put a brand new message filter onto a ZMQ_SUB socket. And here's the tea: freshly-created ZMQ_SUB sockets filter out everything by default, so like... you literally have to call this option to set up an initial message filter. No filter, no messages. Periodt.

  • If your option_value is empty and length is zero — congrats, you're subscribed to all incoming messages. Everything. The whole feed. 👑
  • If your option_value is non-empty, you subscribe to every message that begins with that prefix.

You can stack multiple filters on one single ZMQ_SUB socket, and a message gets accepted if it matches at least one of them. Union energy ✨

  • Option value type — binary data
  • Option value unit — N/A
  • Default value — N/A
  • Applicable socket types — ZMQ_SUB

🧚 ZMQ_TCP_KEEPALIVE: Override SO_KEEPALIVE socket option

Overrides the SO_KEEPALIVE socket option (where the OS actually supports it, iykyk). Default value of -1 means skip the override entirely and let the OS do its thing. 🫶

  • Option value type — int
  • Option value unit — -1,0,1
  • Default value — -1 (leave to OS default)
  • Applicable socket types — all, when using TCP transports.

💅 ZMQ_TCP_KEEPALIVE_CNT: Override TCP_KEEPCNT socket option

Overrides the TCP_KEEPCNT socket option (where supported by the OS, of course). Default -1 = no overrides, OS default reigns supreme. ✨

  • Option value type — int
  • Option value unit — -1,>0
  • Default value — -1 (leave to OS default)
  • Applicable socket types — all, when using TCP transports.

🦋 ZMQ_TCP_KEEPALIVE_IDLE: Override TCP_KEEPIDLE (or TCP_KEEPALIVE on some OS)

Overrides the TCP_KEEPIDLE (or TCP_KEEPALIVE on some OS) socket option — where supported by OS, obviously. Default -1 means skip the overrides, leave it to the OS default. Simple as. 💜

  • Option value type — int
  • Option value unit — -1,>0
  • Default value — -1 (leave to OS default)
  • Applicable socket types — all, when using TCP transports.

🎀 ZMQ_TCP_KEEPALIVE_INTVL: Override TCP_KEEPINTVL socket option

Overrides the TCP_KEEPINTVL socket option (where supported by OS). Default -1 = skip any overrides and leave it to the OS default. We love a hands-off queen. 👑

  • Option value type — int
  • Option value unit — -1,>0
  • Default value — -1 (leave to OS default)
  • Applicable socket types — all, when using TCP transports.

🔥 ZMQ_TCP_MAXRT: Set TCP Maximum Retransmit Timeout

On OSes where it's supported, this sets how long before an unacknowledged TCP retransmit times out. Here's the situation: the system normally attempts many TCP retransmits following an exponential backoff strategy. Which means after a network outage it can take a long time before the session can be re-established. 😭 Setting this option lets that timeout happen at a shorter interval. Chef's kiss for fast recovery.

  • Option value type — int
  • Option value unit — milliseconds
  • Default value — 0 (leave to OS default)
  • Applicable socket types — all, when using TCP transports.

🌈 ZMQ_TOS: Set the Type-of-Service on socket

Sets the ToS fields — that's the Differentiated services (DS) and Explicit Congestion Notification (ECN) field of the IP header. The ToS field is typically used to specify a packet's priority. The availability of this option is dependent on intermediate network equipment that inspects the ToS field and provides a path for low-delay, high-throughput, highly-reliable service, etc. So like... it's giving aspirational, depending on your network. 🙈

  • Option value type — int
  • Option value unit — >0
  • Default value — 0
  • Applicable socket types — all, only for connection-oriented transports

💀 ZMQ_UNSUBSCRIBE: Remove message filter

ZMQ_UNSUBSCRIBE removes an existing message filter from a ZMQ_SUB socket. The filter you specify must match an existing filter that was previously established with ZMQ_SUBSCRIBE — no ghost filters, bestie. And if the socket has several instances of the same filter attached, ZMQ_UNSUBSCRIBE removes only one instance, leaving the rest in place and fully functional. One at a time, no cap. ✨

  • Option value type — binary data
  • Option value unit — N/A
  • Default value — N/A
  • Applicable socket types — ZMQ_SUB

🍓 ZMQ_XPUB_VERBOSE: pass duplicate subscribe messages on XPUB socket

Sets the XPUB socket behaviour when new duplicated subscriptions come in. Enabled? The socket passes all subscribe messages to the caller. Disabled? Only the first subscription to each filter gets passed. Default is 0 (disabled). 🤫

  • Option value type — int
  • Option value unit — 0, 1
  • Default value — 0
  • Applicable socket types — ZMQ_XPUB

⭐ ZMQ_XPUB_VERBOSER: pass duplicate subscribe and unsubscribe messages on XPUB socket

Sets the XPUB socket behaviour on new duplicated subscriptions and unsubscriptions. Enabled? The socket passes all subscribe and unsubscribe messages to the caller. Disabled? Only the first subscription to each filter and the last unsubscription from each filter will be passed. Default is 0 (disabled). It's the verbose option's glow-up fr fr 💅

  • Option value type — int
  • Option value unit — 0, 1
  • Default value — 0
  • Applicable socket types — ZMQ_XPUB

🧠 ZMQ_XPUB_MANUAL: change the subscription handling to manual

Sets the XPUB socket subscription handling mode to manual/automatic. A value of 0 is the default, and subscription requests get handled automatically. A value of 1 flips subscription request handling to manual — and in manual mode, subscription requests are not added to the subscription list. To actually add a subscription, the user needs to call setsockopt with ZMQ_SUBSCRIBE on the XPUB socket. You're driving now, bestie. 🚗

  • Option value type — int
  • Option value unit — 0, 1
  • Default value — 0
  • Applicable socket types — ZMQ_XPUB

🫧 ZMQ_XPUB_MANUAL_LAST_VALUE: change the subscription handling to manual

This one is similar to ZMQ_XPUB_MANUAL, but here's the difference: ZMQ_XPUB_MANUAL_LAST_VALUE changes the XPUB socket behaviour to send the first message to the last subscriber after the socket receives a subscription and calls setsockopt with ZMQ_SUBSCRIBE on the XPUB socket. This prevents duplicated messages when using last value caching (LVC). Very demure, very mindful. 🎀

Note: in DRAFT state, not yet available in stable releases. 😭
  • Option value type — int
  • Option value unit — 0, 1
  • Default value — 0
  • Applicable socket types — ZMQ_XPUB

🚫 ZMQ_XPUB_NODROP: do not silently drop messages if SENDHWM is reached

Sets the XPUB socket behaviour to return error EAGAIN if SENDHWM is reached and the message could not be send. (Yes, that's the original phrasing, we keep it.)

A value of 0 is the default and drops the message silently when the peer's SNDHWM is reached. A value of 1 returns an EAGAIN error code if the SNDHWM is reached and ZMQ_DONTWAIT was used. No more ghosting your messages. 💀

  • Option value type — int
  • Option value unit — 0, 1
  • Default value — 0
  • Applicable socket types — ZMQ_XPUB, ZMQ_PUB

👋 ZMQ_XPUB_WELCOME_MSG: set welcome message that will be received by subscriber when connecting

Sets a welcome message that will be received by a subscriber when connecting. The subscriber must subscribe to the Welcome message before connecting — that's the rule. The welcome message will also be sent on reconnecting. And for the welcome message to work well, the user must poll on incoming subscription messages on the XPUB socket and handle them. It's giving ✨ do the work ✨

Use NULL and length of zero to disable the welcome message.

  • Option value type — binary data
  • Option value unit — N/A
  • Default value — NULL
  • Applicable socket types — ZMQ_XPUB

🦋 ZMQ_XSUB_VERBOSE_UNSUBSCRIBE: pass duplicate unsubscribe messages on XSUB socket

Sets the XSUB socket behaviour on duplicated unsubscriptions. Enabled? The socket passes all unsubscribe messages to the caller. Disabled? Only the last unsubscription from each filter gets passed. Default is 0 (disabled). 🙈

Note: in DRAFT state, not yet available in stable releases.
  • Option value type — int
  • Option value unit — 0, 1
  • Default value — 0
  • Applicable socket types — ZMQ_XSUB

🥇 ZMQ_ONLY_FIRST_SUBSCRIBE: Process only first subscribe/unsubscribe in a multipart message

If set, only the first part of the multipart message is processed as a subscribe/unsubscribe message. The rest are forwarded as user data regardless of message contents. The audacity of parts 2 through n, honestly. 😌

If not set (that's the default), subscribe/unsubscribe messages in a multipart message are processed as such regardless of their number and order.

  • Option value type — int
  • Option value unit — boolean
  • Default value — 0 (false)
  • Applicable socket types — ZMQ_XSUB, ZMQ_XPUB

🔐 ZMQ_ZAP_DOMAIN: Set RFC 27 authentication domain

Sets the domain for ZAP (ZMQ RFC 27) authentication. A ZAP domain must be specified to enable authentication — no domain, no auth. When the ZAP domain is empty, which is the default, ZAP authentication is disabled. Heads up though: this is not compatible with previous versions of libzmq, so it can be controlled by ZMQ_ZAP_ENFORCE_DOMAIN which for now is disabled by default. See http://rfc.zeromq.org/spec:27 for more details. 🌐

  • Option value type — character string
  • Option value unit — N/A
  • Default value — empty
  • Applicable socket types — all, when using TCP transport

📜 ZMQ_ZAP_ENFORCE_DOMAIN: Set ZAP domain handling to strictly adhere the RFC

The ZAP (ZMQ RFC 27) authentication protocol specifies that a domain must always be set. Older versions of libzmq did not follow the spec and allowed an empty domain to be set. 🙃 This option lets you enable or disable the stricter, backward incompatible behaviour. For now it's disabled by default, but in a future version it will be enabled by default. Mark your calendars, bestie.

  • Option value type — int
  • Option value unit — 0, 1
  • Default value — 0
  • Applicable socket types — all, when using ZAP

🕸️ ZMQ_TCP_ACCEPT_FILTER: Assign filters to allow new TCP connections

Assigns an arbitrary number of filters that get applied for each new TCP transport connection on a listening socket. If no filters are applied, the TCP transport allows connections from any IP address. If at least one filter is applied, then the new connection's source IP should be matched. To clear all filters, call zmq_setsockopt(socket, ZMQ_TCP_ACCEPT_FILTER, NULL, 0). A filter is a null-terminated string with ipv6 or ipv4 CIDR. 🧵

Note: This option is deprecated — please use authentication via the ZAP API and IP address allowing / blocking instead. Do not build new things on this, bestie, I'm begging. 😭
  • Option value type — binary data
  • Option value unit — N/A
  • Default value — no filters (allow from all)
  • Applicable socket types — all listening sockets, when using TCP transports.

ZMQ_IPC_FILTER_GID: Assign group ID filters to allow new IPC connections 👑

okay so this one lets you assign an arbitrary number of filters that will be applied for each new IPC transport connection on a listening socket. If no IPC filters are applied, then the IPC transport allows connections from any process. If at least one UID, GID, or PID filter is applied then new connection credentials should be matched. To clear all GID filters call zmq_setsockopt(socket, ZMQ_IPC_FILTER_GID, NULL, 0).

Note — GID filters are only available on platforms supporting SO_PEERCRED or LOCAL_PEERCRED socket options (currently only Linux and later versions of OS X).
Note — This option is deprecated, please use authentication via the ZAP API and IPC allowing / blocking.
  • Option value type — gid_t
  • Option value unit — N/A
  • Default value — no filters (allow from all)
  • Applicable socket types — all listening sockets, when using IPC transports.

ZMQ_IPC_FILTER_PID: Assign process ID filters to allow new IPC connections 🧚

bestie, same vibe here but for process IDs. Assign an arbitrary number of filters that will be applied for each new IPC transport connection on a listening socket. If no IPC filters are applied, then the IPC transport allows connections from any process. If at least one UID, GID, or PID filter is applied then new connection credentials should be matched. To clear all PID filters call zmq_setsockopt(socket, ZMQ_IPC_FILTER_PID, NULL, 0).

Note — PID filters are only available on platforms supporting the SO_PEERCRED socket option (currently only Linux).
Note — This option is deprecated, please use authentication via the ZAP API and IPC allowing / blocking.
  • Option value type — pid_t
  • Option value unit — N/A
  • Default value — no filters (allow from all)
  • Applicable socket types — all listening sockets, when using IPC transports.

ZMQ_IPC_FILTER_UID: Assign user ID filters to allow new IPC connections 💜

Now for user IDs, the way that this works is you assign an arbitrary number of filters that will be applied for each new IPC transport connection on a listening socket. If no IPC filters are applied, then the IPC transport allows connections from any process. If at least one UID, GID, or PID filter is applied then new connection credentials should be matched. To clear all UID filters call zmq_setsockopt(socket, ZMQ_IPC_FILTER_UID, NULL, 0).

Note — UID filters are only available on platforms supporting SO_PEERCRED or LOCAL_PEERCRED socket options (currently only Linux and later versions of OS X).
Note — This option is deprecated, please use authentication via the ZAP API and IPC allowing / blocking.
  • Option value type — uid_t
  • Option value unit — N/A
  • Default value — no filters (allow from all)
  • Applicable socket types — all listening sockets, when using IPC transports.

ZMQ_IPV4ONLY: Use IPv4-only on socket 🌈

Set the IPv4-only option for the socket. This option is deprecated. Please use the ZMQ_IPV6 option.

  • Option value type — int
  • Option value unit — boolean
  • Default value — 1 (true)
  • Applicable socket types — all, when using TCP transports.

ZMQ_VMCI_BUFFER_SIZE: Set buffer size of the VMCI socket 🦋

The ZMQ_VMCI_BUFFER_SIZE option shall set the size of the underlying buffer for the socket. Used during negotiation before the connection is established.

  • Option value type — uint64_t
  • Option value unit — bytes
  • Default value — 65546
  • Applicable socket types — all, when using VMCI transport

ZMQ_VMCI_BUFFER_MIN_SIZE: Set min buffer size of the VMCI socket 🎀

The ZMQ_VMCI_BUFFER_MIN_SIZE option shall set the min size of the underlying buffer for the socket. Used during negotiation before the connection is established.

  • Option value type — uint64_t
  • Option value unit — bytes
  • Default value — 128
  • Applicable socket types — all, when using VMCI transport

ZMQ_VMCI_BUFFER_MAX_SIZE: Set max buffer size of the VMCI socket ✨

The ZMQ_VMCI_BUFFER_MAX_SIZE option shall set the max size of the underlying buffer for the socket. Used during negotiation before the connection is established.

  • Option value type — uint64_t
  • Option value unit — bytes
  • Default value — 262144
  • Applicable socket types — all, when using VMCI transport

ZMQ_VMCI_CONNECT_TIMEOUT: Set connection timeout of the VMCI socket ⏳

The ZMQ_VMCI_CONNECT_TIMEOUT option shall set connection timeout for the socket.

  • Option value type — int
  • Option value unit — milliseconds
  • Default value — -1
  • Applicable socket types — all, when using VMCI transport

ZMQ_MULTICAST_LOOP: Control multicast local loopback 🔁

For multicast UDP sender sockets this option sets whether the data sent should be looped back on local listening sockets.

  • Option value type — int
  • Option value unit — 0, 1
  • Default value — 1
  • Applicable socket types — ZMQ_RADIO, when using UDP multicast transport

ZMQ_ROUTER_NOTIFY: Send connect and disconnect notifications 🔔

Enable connect and disconnect notifications on a ROUTER socket. When enabled, the socket delivers a zero-length message (with routing-id as first frame) when a peer connects or disconnects. It’s possible to notify both events for a peer by OR-ing the flag values. This option only applies to stream oriented (tcp, ipc) transports.

Note — in DRAFT state, not yet available in stable releases.
  • Option value type — int
  • Option value unit — 0, ZMQ_NOTIFY_CONNECT, ZMQ_NOTIFY_DISCONNECT, ZMQ_NOTIFY_CONNECT | ZMQ_NOTIFY_DISCONNECT
  • Default value — 0
  • Applicable socket types — ZMQ_ROUTER

ZMQ_IN_BATCH_SIZE: Maximal receive batch size 📥

Sets the maximal amount of messages that can be received in a single recv system call. WARNING: this option should almost never be changed. The default has been chosen to offer the best compromise between latency and throughtput. In the vast majority of cases, changing this option will result in worst result if not outright breakages.

Cannot be zero.

Note — in DRAFT state, not yet available in stable releases.
  • Option value type — int
  • Option value unit — messages
  • Default value — 8192
  • Applicable socket types — All, when using TCP, IPC, PGM or NORM transport.

ZMQ_OUT_BATCH_SIZE: Maximal send batch size 📤

Sets the maximal amount of messages that can be sent in a single send system call. WARNING: this option should almost never be changed. The default has been chosen to offer the best compromise between latency and throughtput. In the vast majority of cases, changing this option will result in worst result if not outright breakages.

Cannot be zero.

Note — in DRAFT state, not yet available in stable releases.
  • Option value type — int
  • Option value unit — messages
  • Default value — 8192
  • Applicable socket types — All, when using TCP, IPC, PGM or NORM transport.

ZMQ_NORM_MODE: NORM Sender Mode 🚀

Sets the NORM sender mode to control the operation of the NORM transport. NORM supports fixed rate operation (0=ZMQ_NORM_FIXED), congestion control mode (1=ZMQ_NORM_CC), loss-tolerant congestion control (2=ZMQ_NORM_CCL), explicit congestion notification (ECN)-enabled congestion control (3=ZMQ_NORM_CCE), and ECN-only congestion control (4=ZMQ_NORM_CCE_ECNONLY). The default value is TCP-friendly congestion control mode. Fixed rate mode (using datarate set by ZMQ_RATE) offers better performance, but care must be taken to prevent data loss. ECN modes will set one of the ECN Capable Transport bits in the given ZMQ_TOS if it is not set already.

Note — in DRAFT state, not yet available in stable releases.
  • Option value type — int
  • Option value unit — 0, 1, 2, 3, 4
  • Default value — 1 (ZMQ_NORM_CC)
  • Applicable socket types — All, when using NORM transport.

ZMQ_NORM_UNICAST_NACK: Set NORM Unicast NACK mode 📨

If set, NORM receiver will send Negative ACKnowledgements (NACKs) back to the sender using unicast instead of multicast. NORM transport endpoints specifying a unicast address will enable this by default, but it is disabled by default for multicast addresses.

Note — in DRAFT state, not yet available in stable releases.
  • Option value type — int
  • Option value unit — boolean
  • Default value — 0 (false)
  • Applicable socket types — All, when using NORM transport.

ZMQ_NORM_BUFFER_SIZE: Set NORM buffer size 🧺

Sets NORM buffer size for NORM transport sender, receiver, and stream.

Note — in DRAFT state, not yet available in stable releases.
  • Option value type — int
  • Option value unit — kilobytes
  • Default value — 2048
  • Applicable socket types — All, when using NORM transport.

ZMQ_NORM_SEGMENT_SIZE: Set NORM segment size 📏

Sets NORM sender segment size, which is the maximum message payload size of individual NORM messages (ZMQ messages may be split over multiple NORM messages). Ideally, this value should fit within the system/network maximum transmission unit (MTU) after accounting for additional NORM message headers (up to 48 bytes).

Note — in DRAFT state, not yet available in stable releases.
  • Option value type — int
  • Option value unit — bytes
  • Default value — 1400
  • Applicable socket types — All, when using NORM transport.

ZMQ_NORM_BLOCK_SIZE: Set NORM block size 🧱

Sets NORM sender block size, which is the number of segments in a NORM FEC coding block. NORM repair operations take place at block boundaries. Maximum value is 255, but parity packets (ZMQ_NORM_NUM_PARITY) are limited to a value of (255 - ZMQ_NORM_BLOCK_SIZE). Minimum value is (ZMQ_NORM_NUM_PARITY + 1). Effective value may be different based on the settings of ZMQ_NORM_NUM_PARITY and ZMQ_NORM_NUM_AUTOPARITY if invalid settings are provided.

Note — in DRAFT state, not yet available in stable releases.
  • Option value type — int
  • Option value unit — >0, ⟨255
  • Default value — 16
  • Applicable socket types — literally all of them, when you're using the NORM transport.

ZMQ_NORM_NUM_PARITY: Set number of NORM parity segments

Okay bestie, this one sets the maximum number of NORM parity symbol segments that the sender is willing to calculate per FEC coding block, so it can repair lost data. ✨

The maximum value is 255, but it gets further limited to a value of (255 - ZMQ_NORM_BLOCK_SIZE). Minimum value is ZMQ_NORM_NUM_AUTOPARITY. And heads up — the effective value may be different based on the setting of ZMQ_NORM_NUM_AUTOPARITY if you hand it invalid settings. 🙈

Note: in DRAFT state, not yet available in stable releases. 🚧
  • Option value type — int
  • Option value unit — >0, <255
  • Default value — 4
  • Applicable socket types — All, when using NORM transport.

ZMQ_NORM_NUM_AUTOPARITY: Set number of proactive NORM parity segments

This sets the number of NORM parity symbol segments that the sender will proactively send at the end of each FEC coding block. 🎀

By default, no proactive parity segments get sent at all — instead, parity segments only go out in response to repair requests (NACKs). Maximum value is 255, but it's further limited to a maximum value of ZMQ_NORM_NUM_PARITY. So it's giving capped, fr fr.

Note: in DRAFT state, not yet available in stable releases. 🚧
  • Option value type — int
  • Option value unit — >=0, <255
  • Default value — 0
  • Applicable socket types — All, when using NORM transport.

ZMQ_NORM_PUSH: Enable NORM push mode

Enables NORM stream push mode, which alters the behavior of the sender when enqueueing new data. 👑

By default, NORM will stop accepting new messages while it's waiting for old data to be transmitted and/or repaired. Enabling push mode discards the oldest data — which may be pending repair or may never have been transmitted — in favor of accepting new data. So this is useful in cases where it's more important to quickly deliver new data instead of reliably delivering older data. It's giving "new stuff first, idc about the old stuff" energy. 💅

Note: in DRAFT state, not yet available in stable releases. 🚧
  • Option value type — int
  • Option value unit — boolean
  • Default value — 0 (false)
  • Applicable socket types — All, when using NORM transport.

✨ RETURN VALUE ✨

The zmq_setsockopt() function shall return zero if successful. Otherwise it shall return -1 and set errno to one of the values defined below. That's it, that's the deal. 🫶

💀 ERRORS 💀

  • EINVAL — The requested option option_name is unknown, or the requested option_len or option_value is invalid.
  • ETERM — The 0MQ context associated with the specified socket was terminated.
  • ENOTSOCK — The provided socket was invalid.
  • EINTR — The operation was interrupted by delivery of a signal.

🌈 EXAMPLE 🌈

Subscribing to messages on a ZMQ_SUB socket. Classic.

/* Subscribe to all messages */
rc = zmq_setsockopt (socket, ZMQ_SUBSCRIBE, "", 0);
assert (rc == 0);
/* Subscribe to messages prefixed with "ANIMALS.CATS" */
rc = zmq_setsockopt (socket, ZMQ_SUBSCRIBE, "ANIMALS.CATS", 12);

Setting I/O thread affinity. Look at her go 😭

int64_t affinity;
/* Incoming connections on TCP port 5555 shall be handled by I/O thread 1 */
affinity = 1;
rc = zmq_setsockopt (socket, ZMQ_AFFINITY, &affinity, sizeof (affinity));
assert (rc);
rc = zmq_bind (socket, "tcp://lo:5555");
assert (rc);
/* Incoming connections on TCP port 5556 shall be handled by I/O thread 2 */
affinity = 2;
rc = zmq_setsockopt (socket, ZMQ_AFFINITY, &affinity, sizeof (affinity));
assert (rc);
rc = zmq_bind (socket, "tcp://lo:5556");
assert (rc);

👀 SEE ALSO 👀

zmq_getsockopt(3) zmq_socket(3) zmq_plain(7) zmq_curve(7) zmq(7)

🎀 AUTHORS 🎀

This page was written by the 0MQ community. To make a change please read the 0MQ Contribution Policy at http://www.zeromq.org/docs:contributing. The community ate, no cap. 💖

And that's the whole page bestie, we made it to the end!! 🌸 This was 0MQ 4.3.5, dated 10/09/2023, and yes the page is literally called ZMQ_SETSOCKOPT(3) — go forth and set your socket options responsibly, mwah 💅✨

📦 where this page comes from