pub struct Config {Show 27 fields
pub deliver_subject: Option<String>,
pub durable_name: Option<String>,
pub name: Option<String>,
pub description: Option<String>,
pub deliver_group: Option<String>,
pub deliver_policy: DeliverPolicy,
pub ack_policy: AckPolicy,
pub ack_wait: Duration,
pub max_deliver: i64,
pub filter_subject: String,
pub filter_subjects: Vec<String>,
pub replay_policy: ReplayPolicy,
pub rate_limit: u64,
pub sample_frequency: u8,
pub max_waiting: i64,
pub max_ack_pending: i64,
pub headers_only: bool,
pub flow_control: bool,
pub idle_heartbeat: Duration,
pub max_batch: i64,
pub max_bytes: i64,
pub max_expires: Duration,
pub inactive_threshold: Duration,
pub num_replicas: usize,
pub memory_storage: bool,
pub metadata: HashMap<String, String>,
pub backoff: Vec<Duration>,
}
Expand description
Configuration for consumers. From a high level, the
durable_name
and deliver_subject
fields have a particularly
strong influence on the consumer’s overall behavior.
Fields§
§deliver_subject: Option<String>
Setting deliver_subject
to Some(...)
will cause this consumer
to be “push-based”. This is analogous in some ways to a normal
NATS subscription (rather than a queue subscriber) in that the
consumer will receive all messages published to the stream that
the consumer is interested in. Acknowledgment policies such as
AckPolicy::None
and AckPolicy::All
may be enabled for such
push-based consumers, which reduce the amount of effort spent
tracking delivery. Combining AckPolicy::All
with
Consumer::process_batch
enables particularly nice throughput
optimizations.
Setting deliver_subject
to None
will cause this consumer to
be “pull-based”, and will require explicit acknowledgment of
each message. This is analogous in some ways to a normal NATS
queue subscriber, where a message will be delivered to a single
subscriber. Pull-based consumers are intended to be used for
workloads where it is desirable to have a single process receive
a message. The only valid ack_policy
for pull-based consumers
is the default of AckPolicy::Explicit
, which acknowledges each
processed message individually. Pull-based consumers may be a
good choice for work queue-like workloads where you want messages
to be handled by a single consumer process. Note that it is
possible to deliver a message to multiple consumers if the
consumer crashes or is slow to acknowledge the delivered message.
This is a fundamental behavior present in all distributed systems
that attempt redelivery when a consumer fails to acknowledge a message.
This is known as “at least once” message processing. To achieve
“exactly once” semantics, it is necessary to implement idempotent
semantics in any system that is written to as a result of processing
a message.
durable_name: Option<String>
Setting durable_name
to Some(...)
will cause this consumer
to be “durable”. This may be a good choice for workloads that
benefit from the JetStream
server or cluster remembering the
progress of consumers for fault tolerance purposes. If a consumer
crashes, the JetStream
server or cluster will remember which
messages the consumer acknowledged. When the consumer recovers,
this information will allow the consumer to resume processing
where it left off. If you’re unsure, set this to Some(...)
.
Setting durable_name
to None
will cause this consumer to
be “ephemeral”. This may be a good choice for workloads where
you don’t need the JetStream
server to remember the consumer’s
progress in the case of a crash, such as certain “high churn”
workloads or workloads where a crashed instance is not required
to recover.
name: Option<String>
A name of the consumer. Can be specified for both durable and ephemeral consumers.
description: Option<String>
A short description of the purpose of this consumer.
deliver_group: Option<String>
Deliver group to use.
deliver_policy: DeliverPolicy
Allows for a variety of options that determine how this consumer will receive messages
ack_policy: AckPolicy
How messages should be acknowledged
ack_wait: Duration
How long to allow messages to remain un-acknowledged before attempting redelivery
max_deliver: i64
Maximum number of times a specific message will be delivered. Use this to avoid poison pill messages that repeatedly crash your consumer processes forever.
filter_subject: String
When consuming from a Stream with many subjects, or wildcards, this selects only specific incoming subjects. Supports wildcards.
filter_subjects: Vec<String>
server_2_10
only.Fulfills the same role as Config::filter_subject, but allows filtering by many subjects.
replay_policy: ReplayPolicy
Whether messages are sent as quickly as possible or at the rate of receipt
rate_limit: u64
The rate of message delivery in bits per second
sample_frequency: u8
What percentage of acknowledgments should be samples for observability, 0-100
max_waiting: i64
The maximum number of waiting consumers.
max_ack_pending: i64
The maximum number of unacknowledged messages that may be in-flight before pausing sending additional messages to this consumer.
headers_only: bool
Only deliver headers without payloads.
flow_control: bool
Enable flow control messages
idle_heartbeat: Duration
Enable idle heartbeat messages
max_batch: i64
Maximum size of a request batch
max_bytes: i64
Maximum size of a request max_bytes
max_expires: Duration
Maximum value for request expiration
inactive_threshold: Duration
Threshold for ephemeral consumer inactivity
num_replicas: usize
Number of consumer replicas
memory_storage: bool
Force consumer to use memory storage.
metadata: HashMap<String, String>
server_2_10
only.Additional consumer metadata.
backoff: Vec<Duration>
Custom backoff for missed acknowledgments.