package pulse.config.outflow.v1
Get desktop application:
View/edit binary Protocol Buffers messages
message
wire.proto:17Common configuration for wire clients.
Used in: , , ,
string send_to = 1
Address to send to. IP/port for TCP/UDP clients or name for unix domain sockets. Ignored for the null outflow.
optional protocol = 2
Protocol type.
optional queue_policy = 3
Queue policy for the client.
optional uint64 batch_max_bytes = 4
The maximum numer of bytes per batch. Defaults to 8KiB.
optional write_timeout = 5
How long to wait for a write to complete before considering the socket to be failed. Defaults to no timeout.
optional uint64 max_in_flight = 6
The maximum number of inflight requests. Defaults to 16.
message
outflow_common.proto:15Authentication configuration for Prometheus remote write.
Used in: ,
oneof auth_type
bearer_token = 2
Bearer token configuration.
aws = 3
AWS auth configuration.
message
outflow_common.proto:17Use AWS authentication.
Used in:
oneof auth_type
default = 1
Default configuration.
message
outflow_common.proto:20Use the default credential chain, region provider, and SDK configuration specified by the environment.
Used in:
optional string assume_role = 1
optional string signing_region = 2
Optional signing region override. Otherwise, the region will be determined by the default region provider chain.
message
wire.proto:52Configuration for a client that writes to the equivalent of /dev/null.
Used in:
optional common = 1
Common wire client configuration.
message
otlp.proto:18Used in:
string send_to = 1
The URI to send to.
optional request_timeout = 2
The request timeout. Defaults to 10s.
optional connect_timeout = 11
The connection timeout for establishing a new upstream connection. Defaults to 250ms.
optional uint64 max_in_flight = 3
The maximum number of inflight requests. Defaults to 16.
optional queue_policy = 4
Queue policy for the client.
optional uint64 batch_max_samples = 5
The maximum numer of samples per batch. Defaults to 1000.
optional auth = 6
What auth to use. If not specified, no auth is used.
repeated request_headers = 7
Optional additional request headers that are sent with the remote write request.
optional retry_policy = 8
Retry policy. If not specified default values are used.
OtlpClientConfig.OtlpCompression compression = 9
Compression to use. Defaults to SNAPPY.
bool convert_names_to_prometheus = 10
If sending OTLP to a backend that expects Prometheus-style names, convert the metric names to Prometheus format: https://prometheus.io/docs/concepts/data_model/#metric-names-and-labels For tags invalid characters will be replaced with '_'. For names, '.' will be replaced with ':' and other invalid characters will be replaced with '_'.
Used in:
SNAPPY = 0
NONE = 1
message
outflow.proto:17Configuration for an individual outflow.
Used in:
oneof config_type
null_outflow = 1
unix = 2
tcp = 3
udp = 4
prom_remote_write = 5
otlp = 6
message
prom_remote_write.proto:19Configuration for a Prometheus remote write outflow/client.
Used in: ,
string send_to = 1
The URI to send to.
optional request_timeout = 2
The request timeout. Defaults to 10s.
optional connect_timeout = 14
The connection timeout for establishing a new upstream connection. Defaults to 250ms.
optional uint64 max_in_flight = 3
The maximum number of inflight requests. Defaults to 16.
optional queue_policy = 4
Queue policy for the client.
optional uint64 batch_max_samples = 5
The maximum number of samples per batch. Defaults to 1000.
optional uint64 batch_max_size = 13
The maximum size of each batch in bytes, *after* compression. There is no default for this. This can be used of the TSDB has a strict maximum request size. Since it's impossible to know the size before compression, batch_max_samples should be tuned alongside this. If the limit is hit, the batch will be split in half into 2 requests so this is an expensive operation.
bool metadata_only = 6
Whether to send metadata (name and tags) only. If true, samples are dropped.
optional auth = 7
What auth to use. If not specified, no auth is used.
repeated request_headers = 8
Optional additional request headers that are sent with the remote write request.
optional bool convert_metric_name = 9
Whether to attempt conversion to Prometheus name format: https://prometheus.io/docs/concepts/data_model/#metric-names-and-labels. If false, the name will be passed through even if it is not compliant. Defaults to true.
optional retry_policy = 10
Retry policy. If not specified default values are used.
optional lyft_specific_config = 11
Lyft specific configuration. To be generalized later.
optional pool_idle_timeout = 12
Idle time for outbound connections in the connection pool. If not set, defaults to 90s.
message
prom_remote_write.proto:23Configuration specific to the Lyft's use of M3. This can be generalized later if needed. Will end up setting `M3-Metrics-Type: aggregated` and `M3-Storage-Policy` to the value of each policy below.
Used in:
string general_storage_policy = 1
Catch-all storage policy.
optional string instance_metrics_storage_policy = 2
Metrics that have a "host" tag.
optional string cloudwatch_metrics_storage_policy = 3
Metrics that do not have a "statsd" "source" tag, and are of the form <namespace>:infra:aws.
message
queue_policy.proto:18Defines the policy used for batching and queueing data within clients. In general memory usage is mostly bounded (some of the byte calculations are fuzzy when dealing with proto outputs) by the limits specified within this policy. TODO(mattkleni123): Add a drop policy, including queueing to disk.
Used in: , ,
optional uint64 queue_max_bytes = 1
The maximum number of bytes that can be queued for sending before dropping. Bytes beyond this are dropped. Note that this tracks only data that is queued to send. It does not include any pending batch data. Defaults to 8MiB.
optional batch_fill_wait = 2
The wait period between batch fills. Batches will be sent by this deadline if not full. Defaults to 50ms.
optional uint64 concurrent_batch_queues = 3
The number of concurrent pending batches to maintain. In a high throughput environment it is possible for the pending batch to become a point of contention, especially for Prometheus remote write where this is significant computation required to complete a batch (proto encoding and compression). By increasing concurrency it is more likely that CPU can become saturated. Currently this must be set manually and should be tuned depending on the environment. A good setting for a high throughput scenario is likely somewhere between CPU requests and limit. If not set, this default to 1.
bool use_fifo = 4
By default the queue operates as a LIFO queue to prioritize newer data. If set to true, the queue operates as a FIFO queue. This may be desirable in some scenarios to preserve ordering of data. Defaults to false.
message
outflow_common.proto:48An individual request header sent with the remote write request.
Used in: ,
string name = 1
The name of the header.
oneof value_type
value = 2
The fixed value of the header.
bool timestamp = 3
Replaced with the timestamp, defined as the number of seconds since the unix epoch.
message
wire.proto:58Configuration for a client using a TCP socket.
Used in: ,
optional common = 1
Common wire client configuration.
message
wire.proto:46Configuration for a client using UDP.
Used in:
optional common = 1
Common wire client configuration.
message
wire.proto:40Configuration for a client using a unix domain socket.
Used in:
optional common = 1
Common wire client configuration.