zmq_ctx_set - Man Page

By default the context will block, forever, on a zmq_ctx_term call. The assumption behind this behavior is that abrupt termination will cause message loss. Most real applications use some form of handshaking to ensure applications receive termination messages, and then terminate the context with ZMQ_LINGER set to zero on all sockets. This setting is an easier way to get the same result. When ZMQ_BLOCKY is set to false, all new sockets are given a linger timeout of zero. You must still close all sockets before calling zmq_ctx_term.

The ZMQ_IO_THREADS argument specifies the size of the 0MQ thread pool to handle I/O operations. If your application is using only the inproc transport for messaging you may set this to zero, otherwise set it to at least one. This option only applies before creating any sockets on the context.

The ZMQ_THREAD_SCHED_POLICY argument sets the scheduling policy for internal context’s thread pool. This option is not available on windows. Supported values for this option can be found in sched.h file, or at http://man7.org/linux/man-pages/man2/sched_setscheduler.2.html. This option only applies before creating any sockets on the context.

The ZMQ_THREAD_PRIORITY argument sets scheduling priority for internal context’s thread pool. This option is not available on windows. Supported values for this option depend on chosen scheduling policy. On Linux, when the scheduler policy is SCHED_OTHER, SCHED_IDLE or SCHED_BATCH, the OS scheduler will not use the thread priority but rather the thread "nice value"; in such cases, if ZMQ_THREAD_PRIORITY is set to a strictly positive value, the system call "nice" will be used to set the nice value to -20 (max priority) instead of adjusting the thread priority (which must be zero for those scheduling policies). Details can be found in sched.h file, or at http://man7.org/linux/man-pages/man2/sched_setscheduler.2.html. This option only applies before creating any sockets on the context.

The ZMQ_THREAD_AFFINITY_CPU_ADD argument adds a specific CPU to the affinity list for the internal context’s thread pool. This option is only supported on Linux. This option only applies before creating any sockets on the context. The default affinity list is empty and means that no explicit CPU-affinity will be set on internal context’s threads.

The ZMQ_THREAD_AFFINITY_CPU_REMOVE argument removes a specific CPU to the affinity list for the internal context’s thread pool. This option is only supported on Linux. This option only applies before creating any sockets on the context. The default affinity list is empty and means that no explicit CPU-affinity will be set on internal context’s threads.

The ZMQ_THREAD_NAME_PREFIX argument sets a numeric prefix to each thread created for the internal context’s thread pool. This option is only supported on Linux. This option is useful to help debugging done via "top -H" or "gdb"; in case multiple processes on the system are using ZeroMQ it is useful to provide through this context option an application-specific prefix to distinguish ZeroMQ background threads that belong to different processes. This option only applies before creating any sockets on the context.

The ZMQ_MAX_MSGSZ argument sets the maximum allowed size of a message sent in the context. You can query the maximal allowed value with zmq_ctx_get(3) using the ZMQ_MAX_MSGSZ option.

The ZMQ_ZERO_COPY_RECV argument specifies whether the message decoder should use a zero copy strategy when receiving messages. The zero copy strategy can lead to increased memory usage in some cases. This option allows you to use the older copying strategy. You can query the value of this option with zmq_ctx_get(3) using the ZMQ_ZERO_COPY_RECV option. NOTE: in DRAFT state, not yet available in stable releases.

The ZMQ_MAX_SOCKETS argument sets the maximum number of sockets allowed on the context. You can query the maximal allowed value with zmq_ctx_get(3) using the ZMQ_SOCKET_LIMIT option.

The ZMQ_IPV6 argument sets the IPv6 value for all sockets created in the context from this point onwards. A value of 1 means IPv6 is enabled, while 0 means the socket will use only IPv4. When IPv6 is enabled, a socket will connect to, or accept connections from, both IPv4 and IPv6 hosts.