ryerraguntla commented on code in PR #2933: URL: https://github.com/apache/iggy/pull/2933#discussion_r2943400470
########## core/connectors/sinks/influxdb_sink/src/lib.rs: ########## @@ -0,0 +1,774 @@ +/* Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +use async_trait::async_trait; +use base64::{Engine as _, engine::general_purpose}; +use humantime::Duration as HumanDuration; +use iggy_connector_sdk::{ + ConsumedMessage, Error, MessagesMetadata, Sink, TopicMetadata, sink_connector, +}; +use rand::RngExt; +use reqwest::{Client, StatusCode, Url}; +use serde::{Deserialize, Serialize}; +use std::str::FromStr; +use std::sync::Arc; +use std::sync::atomic::{AtomicU32, Ordering}; +use std::time::Duration; +use std::time::SystemTime; +use std::time::UNIX_EPOCH; +use tokio::sync::Mutex; +use tracing::{debug, error, info, warn}; + +sink_connector!(InfluxDbSink); + +const DEFAULT_MAX_RETRIES: u32 = 3; +const DEFAULT_RETRY_DELAY: &str = "1s"; +const DEFAULT_TIMEOUT: &str = "30s"; +const DEFAULT_PRECISION: &str = "us"; +// Maximum attempts for open() connectivity retries +const DEFAULT_MAX_OPEN_RETRIES: u32 = 10; +// Cap for exponential backoff in open() — never wait longer than this +const DEFAULT_OPEN_RETRY_MAX_DELAY: &str = "60s"; +// How many consecutive batch failures open the circuit breaker +const DEFAULT_CIRCUIT_BREAKER_THRESHOLD: u32 = 5; +// How long the circuit stays open before allowing a probe attempt +const DEFAULT_CIRCUIT_COOL_DOWN: &str = "30s"; + +// --------------------------------------------------------------------------- +// Simple consecutive-failure circuit breaker +// --------------------------------------------------------------------------- +#[derive(Debug)] +struct CircuitBreaker { + threshold: u32, + consecutive_failures: AtomicU32, + open_until: Mutex<Option<tokio::time::Instant>>, + cool_down: Duration, +} + +impl CircuitBreaker { + fn new(threshold: u32, cool_down: Duration) -> Self { + CircuitBreaker { + threshold, + consecutive_failures: AtomicU32::new(0), + open_until: Mutex::new(None), + cool_down, + } + } + + /// Call when a batch write succeeds — resets failure count and closes circuit. + fn record_success(&self) { + self.consecutive_failures.store(0, Ordering::SeqCst); + } + + /// Call when a batch write fails after all retries — may open the circuit. + async fn record_failure(&self) { + let failures = self.consecutive_failures.fetch_add(1, Ordering::SeqCst) + 1; + if failures >= self.threshold { + let mut guard = self.open_until.lock().await; + let deadline = tokio::time::Instant::now() + self.cool_down; + *guard = Some(deadline); + warn!( + "Circuit breaker OPENED after {failures} consecutive batch failures. \ + Pausing writes for {:?}.", + self.cool_down + ); + } + } + + /// Returns true if the circuit is currently open (writes should be skipped). + async fn is_open(&self) -> bool { + let mut guard = self.open_until.lock().await; + if let Some(deadline) = *guard { + if tokio::time::Instant::now() < deadline { + return true; + } + // Cool-down elapsed — allow one probe attempt (half-open state) + *guard = None; + self.consecutive_failures.store(0, Ordering::SeqCst); + info!("Circuit breaker entering HALF-OPEN state — probing InfluxDB."); + } + false + } +} + +// --------------------------------------------------------------------------- +// Main connector structs +// --------------------------------------------------------------------------- + +#[derive(Debug)] +pub struct InfluxDbSink { + pub id: u32, + config: InfluxDbSinkConfig, + client: Option<Client>, + state: Mutex<State>, + verbose: bool, + retry_delay: Duration, + circuit_breaker: Arc<CircuitBreaker>, +} + +#[derive(Debug, Clone, Serialize, Deserialize)] +pub struct InfluxDbSinkConfig { + pub url: String, + pub org: String, + pub bucket: String, + pub token: String, + pub measurement: Option<String>, + pub precision: Option<String>, + pub batch_size: Option<u32>, + pub include_metadata: Option<bool>, + pub include_checksum: Option<bool>, + pub include_origin_timestamp: Option<bool>, + pub include_stream_tag: Option<bool>, + pub include_topic_tag: Option<bool>, + pub include_partition_tag: Option<bool>, + pub payload_format: Option<String>, + pub verbose_logging: Option<bool>, + pub max_retries: Option<u32>, + pub retry_delay: Option<String>, + pub timeout: Option<String>, + // How many times open() will retry before giving up + pub max_open_retries: Option<u32>, + // Upper cap on open() backoff delay + pub open_retry_max_delay: Option<String>, + // Circuit breaker configuration + pub circuit_breaker_threshold: Option<u32>, + pub circuit_breaker_cool_down: Option<String>, +} + +#[derive(Debug, Clone, Copy, PartialEq, Eq, Default)] +enum PayloadFormat { + #[default] + Json, + Text, + Base64, +} + +impl PayloadFormat { + fn from_config(value: Option<&str>) -> Self { + match value.map(|v| v.to_ascii_lowercase()).as_deref() { + Some("text") | Some("utf8") => PayloadFormat::Text, + Some("base64") | Some("raw") => PayloadFormat::Base64, + _ => PayloadFormat::Json, + } + } +} + +#[derive(Debug)] +struct State { + messages_processed: u64, + write_errors: u64, +} + +// --------------------------------------------------------------------------- +// Helpers +// --------------------------------------------------------------------------- + +fn parse_duration(value: Option<&str>, default_value: &str) -> Duration { + let raw = value.unwrap_or(default_value); + HumanDuration::from_str(raw) + .map(|d| d.into()) + .unwrap_or_else(|_| Duration::from_secs(1)) +} + +fn is_transient_status(status: StatusCode) -> bool { + status == StatusCode::TOO_MANY_REQUESTS || status.is_server_error() +} + +// Apply ±20% random jitter to a duration to spread retry storms +fn jitter(base: Duration) -> Duration { + let millis = base.as_millis() as u64; + let jitter_range = millis / 5; // 20% of base + if jitter_range == 0 { + return base; + } + let delta = rand::rng().random_range(0..=jitter_range * 2); + Duration::from_millis(millis.saturating_sub(jitter_range) + delta) +} + +// True exponential backoff: base * 2^attempt, capped at max_delay +fn exponential_backoff(base: Duration, attempt: u32, max_delay: Duration) -> Duration { + let factor = 2u64.saturating_pow(attempt); + let raw = Duration::from_millis(base.as_millis().saturating_mul(factor as u128) as u64); + raw.min(max_delay) +} + +// Parse Retry-After header value (integer seconds or HTTP date) +fn parse_retry_after(value: &str) -> Option<Duration> { + if let Ok(secs) = value.trim().parse::<u64>() { + return Some(Duration::from_secs(secs)); + } + // HTTP-date fallback would require httpdate crate; return None to use own backoff + None +} + +fn escape_measurement(value: &str) -> String { + value + .replace('\\', "\\\\") + .replace(',', "\\,") + .replace(' ', "\\ ") +} + +fn escape_tag_value(value: &str) -> String { + value + .replace('\\', "\\\\") + .replace(',', "\\,") + .replace('=', "\\=") + .replace(' ', "\\ ") +} + +fn escape_field_string(value: &str) -> String { + value.replace('\\', "\\\\").replace('"', "\\\"") +} + +// --------------------------------------------------------------------------- +// InfluxDbSink implementation +// --------------------------------------------------------------------------- + +impl InfluxDbSink { + pub fn new(id: u32, config: InfluxDbSinkConfig) -> Self { + let verbose = config.verbose_logging.unwrap_or(false); + let retry_delay = parse_duration(config.retry_delay.as_deref(), DEFAULT_RETRY_DELAY); + + // Build circuit breaker from config + let cb_threshold = config + .circuit_breaker_threshold + .unwrap_or(DEFAULT_CIRCUIT_BREAKER_THRESHOLD); + let cb_cool_down = parse_duration( + config.circuit_breaker_cool_down.as_deref(), + DEFAULT_CIRCUIT_COOL_DOWN, + ); + + InfluxDbSink { + id, + config, + client: None, + state: Mutex::new(State { + messages_processed: 0, + write_errors: 0, + }), + verbose, + retry_delay, + circuit_breaker: Arc::new(CircuitBreaker::new(cb_threshold, cb_cool_down)), + } + } + + fn build_client(&self) -> Result<Client, Error> { + let timeout = parse_duration(self.config.timeout.as_deref(), DEFAULT_TIMEOUT); + Client::builder() + .timeout(timeout) + .build() + .map_err(|e| Error::InitError(format!("Failed to create HTTP client: {e}"))) + } + + fn build_write_url(&self) -> Result<Url, Error> { + let base = self.config.url.trim_end_matches('/'); + let mut url = Url::parse(&format!("{base}/api/v2/write")) + .map_err(|e| Error::InvalidConfigValue(format!("Invalid InfluxDB URL: {e}")))?; + + let precision = self + .config + .precision + .as_deref() + .unwrap_or(DEFAULT_PRECISION); + url.query_pairs_mut() + .append_pair("org", &self.config.org) + .append_pair("bucket", &self.config.bucket) + .append_pair("precision", precision); + + Ok(url) + } + + fn build_health_url(&self) -> Result<Url, Error> { + let base = self.config.url.trim_end_matches('/'); + Url::parse(&format!("{base}/health")) + .map_err(|e| Error::InvalidConfigValue(format!("Invalid InfluxDB URL: {e}"))) + } + + async fn check_connectivity(&self) -> Result<(), Error> { + let client = self.get_client()?; + let url = self.build_health_url()?; + + let response = client + .get(url) + .send() + .await + .map_err(|e| Error::Connection(format!("InfluxDB health check failed: {e}")))?; + + if !response.status().is_success() { + let status = response.status(); + let body = response + .text() + .await + .unwrap_or_else(|_| "failed to read response body".to_string()); + return Err(Error::Connection(format!( + "InfluxDB health check returned status {status}: {body}" + ))); + } + + Ok(()) + } + + // Retry connectivity check with exponential backoff + jitter + // instead of failing hard on the first attempt. + async fn check_connectivity_with_retry(&self) -> Result<(), Error> { + let max_open_retries = self + .config + .max_open_retries + .unwrap_or(DEFAULT_MAX_OPEN_RETRIES) + .max(1); + + let max_delay = parse_duration( + self.config.open_retry_max_delay.as_deref(), + DEFAULT_OPEN_RETRY_MAX_DELAY, + ); + + let mut attempt = 0u32; + loop { + match self.check_connectivity().await { + Ok(()) => { + if attempt > 0 { + info!( + "InfluxDB connectivity established after {attempt} retries \ + for sink connector ID: {}", + self.id + ); + } + return Ok(()); + } + Err(e) => { + attempt += 1; + if attempt >= max_open_retries { + error!( + "InfluxDB connectivity check failed after {attempt} attempts \ + for sink connector ID: {}. Giving up: {e}", + self.id + ); + return Err(e); + } + // Exponential backoff, with jitter + let backoff = jitter(exponential_backoff(self.retry_delay, attempt, max_delay)); + warn!( + "InfluxDB health check failed (attempt {attempt}/{max_open_retries}) \ + for sink connector ID: {}. Retrying in {backoff:?}: {e}", + self.id + ); + tokio::time::sleep(backoff).await; + } + } + } + } + + fn get_client(&self) -> Result<&Client, Error> { + self.client + .as_ref() + .ok_or_else(|| Error::Connection("InfluxDB client is not initialized".to_string())) + } + + fn measurement(&self) -> &str { + self.config + .measurement + .as_deref() + .unwrap_or("iggy_messages") + } + + fn payload_format(&self) -> PayloadFormat { + PayloadFormat::from_config(self.config.payload_format.as_deref()) + } + + fn timestamp_precision(&self) -> &str { + self.config + .precision + .as_deref() + .unwrap_or(DEFAULT_PRECISION) + } + + fn get_max_retries(&self) -> u32 { + self.config + .max_retries + .unwrap_or(DEFAULT_MAX_RETRIES) + .max(1) + } + + fn to_precision_timestamp(&self, micros: u64) -> u64 { + match self.timestamp_precision() { + "ns" => micros.saturating_mul(1_000), + "us" => micros, + "ms" => micros / 1_000, + "s" => micros / 1_000_000, + _ => micros, + } + } + fn line_from_message( + &self, + topic_metadata: &TopicMetadata, + messages_metadata: &MessagesMetadata, + message: &ConsumedMessage, + ) -> Result<String, Error> { + let include_metadata = self.config.include_metadata.unwrap_or(true); + let include_checksum = self.config.include_checksum.unwrap_or(true); + let include_origin_timestamp = self.config.include_origin_timestamp.unwrap_or(true); + + let mut tags = Vec::new(); + if include_metadata && self.config.include_stream_tag.unwrap_or(true) { + tags.push(format!( + "stream={}", + escape_tag_value(&topic_metadata.stream) + )); + } + if include_metadata && self.config.include_topic_tag.unwrap_or(true) { + tags.push(format!("topic={}", escape_tag_value(&topic_metadata.topic))); + } + if include_metadata && self.config.include_partition_tag.unwrap_or(true) { + tags.push(format!("partition={}", messages_metadata.partition_id)); + } + + let mut fields = vec![ + format!( + "message_id=\"{}\"", + escape_field_string(&message.id.to_string()) + ), + format!("offset={}i", message.offset as i64), + ]; + + if include_metadata && !self.config.include_stream_tag.unwrap_or(true) { + fields.push(format!( + "iggy_stream=\"{}\"", + escape_field_string(&topic_metadata.stream) + )); + } + if include_metadata && !self.config.include_topic_tag.unwrap_or(true) { + fields.push(format!( + "iggy_topic=\"{}\"", + escape_field_string(&topic_metadata.topic) + )); + } + if include_metadata && !self.config.include_partition_tag.unwrap_or(true) { + fields.push(format!( + "iggy_partition={}", + messages_metadata.partition_id as i64 + )); + } + + if include_checksum { + fields.push(format!("iggy_checksum={}", message.checksum as i64)); + } + if include_origin_timestamp { + fields.push(format!( + "iggy_origin_timestamp={}", + message.origin_timestamp as i64 + )); + } + + let payload_bytes = message.payload.clone().try_into_vec().map_err(|e| { + Error::CannotStoreData(format!("Failed to convert payload to bytes: {e}")) + })?; + + match self.payload_format() { + PayloadFormat::Json => { + let value: serde_json::Value = + serde_json::from_slice(&payload_bytes).map_err(|e| { + Error::CannotStoreData(format!( + "Payload format is json but payload is invalid JSON: {e}" + )) + })?; + let compact = serde_json::to_string(&value).map_err(|e| { + Error::CannotStoreData(format!("Failed to serialize JSON payload: {e}")) + })?; + fields.push(format!( + "payload_json=\"{}\"", + escape_field_string(&compact) + )); + } + PayloadFormat::Text => { + let text = String::from_utf8(payload_bytes).map_err(|e| { + Error::CannotStoreData(format!( + "Payload format is text but payload is invalid UTF-8: {e}" + )) + })?; + fields.push(format!("payload_text=\"{}\"", escape_field_string(&text))); + } + PayloadFormat::Base64 => { + let encoded = general_purpose::STANDARD.encode(payload_bytes); + fields.push(format!( + "payload_base64=\"{}\"", + escape_field_string(&encoded) + )); + } + } + + let measurement = escape_measurement(self.measurement()); + let tags_fragment = if tags.is_empty() { + String::new() + } else { + format!(",{}", tags.join(",")) + }; + + // message.timestamp is microseconds since Unix epoch. + // If it is 0 (unset by the producer), fall back to now() so points are + // not stored at Unix epoch (year 1970), which falls outside every + // range(start: -1h) query window. + // We also blend the message offset as sub-microsecond nanoseconds so + // that multiple messages in the same batch get distinct timestamps and + // are not deduplicated by InfluxDB (same measurement+tags+time = 1 row). + let base_micros = if message.timestamp == 0 { Review Comment: Implemented the uniqueness with InfluxDB idiomatic offset as tags concept. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
