gitdataai/libs/git/hook/pool/redis.rs
ZhenYi 1fed9fc8ab fix(git/hook): address review findings — fs blocking, redis timeout, backoff, slog
- sync/mod.rs: wrap scan_skills_from_dir in spawn_blocking to avoid
  blocking the async executor; use to_path_buf() to get owned PathBuf
- pool/worker.rs: replace 500ms poll sleep with cancellation-is_cancelled
  check (eliminates artificial latency); add exponential backoff on Redis
  errors (1s base, 32s cap, reset on success)
- pool/redis.rs: add 5s timeout on pool.get() for all three methods
  (next, ack_raw, nak_with_retry) to prevent indefinite blocking on
  unresponsive Redis
- sync/gc.rs: add comment explaining why git gc --auto non-zero exit
  is benign
- webhook_dispatch.rs: remove unnecessary format! wrappers in slog macros
- config/hook.rs: document max_concurrent intent (K8s operator/HPA, not
  the single-threaded worker itself)
2026-04-17 13:20:31 +08:00

177 lines
6.0 KiB
Rust

use crate::error::GitError;
use crate::hook::pool::types::HookTask;
use deadpool_redis::cluster::Connection as RedisConn;
use slog::Logger;
use std::time::Duration;
/// Redis List consumer using BLMOVE for atomic move-from-queue-to-work pattern.
pub struct RedisConsumer {
pool: deadpool_redis::cluster::Pool,
/// Hash-tag-prefixed key prefix, e.g. "{hook}".
prefix: String,
block_timeout_secs: u64,
logger: Logger,
}
const POOL_GET_TIMEOUT: Duration = Duration::from_secs(5);
impl RedisConsumer {
pub fn new(
pool: deadpool_redis::cluster::Pool,
prefix: String,
block_timeout_secs: u64,
logger: Logger,
) -> Self {
Self {
pool,
prefix,
block_timeout_secs,
logger,
}
}
/// Atomically moves a task from the main queue to the work queue using BLMOVE.
/// Blocks up to `block_timeout_secs` waiting for a task.
///
/// Returns `Some((HookTask, task_json))` where `task_json` is the raw JSON string
/// needed for LREM on ACK. Returns `None` if the blocking timed out.
pub async fn next(&self, task_type: &str) -> Result<Option<(HookTask, String)>, GitError> {
let queue_key = format!("{}:{}", self.prefix, task_type);
let work_key = format!("{}:{}:work", self.prefix, task_type);
let redis = tokio::time::timeout(POOL_GET_TIMEOUT, self.pool.get())
.await
.map_err(|_| GitError::Internal("redis pool get timed out".into()))?
.map_err(|e| GitError::Internal(format!("redis pool get failed: {}", e)))?;
let mut conn: RedisConn = redis;
// BLMOVE source destination <LEFT|RIGHT> <LEFT|RIGHT> timeout
let task_json: Option<String> = redis::cmd("BLMOVE")
.arg(&queue_key)
.arg(&work_key)
.arg("RIGHT")
.arg("LEFT")
.arg(self.block_timeout_secs)
.query_async(&mut conn)
.await
.map_err(|e| GitError::Internal(format!("BLMOVE failed: {}", e)))?;
match task_json {
Some(json) => {
match serde_json::from_str::<HookTask>(&json) {
Ok(task) => {
slog::debug!(self.logger, "task dequeued";
"task_id" => %task.id,
"task_type" => %task.task_type,
"queue" => %queue_key
);
Ok(Some((task, json)))
}
Err(e) => {
// Malformed task — remove from work queue and discard
slog::warn!(self.logger, "malformed task JSON, discarding";
"error" => %e,
"queue" => %work_key
);
let _ = self.ack_raw(&work_key, &json).await;
Ok(None)
}
}
}
None => {
// Timed out, no task available
Ok(None)
}
}
}
/// Acknowledge a task: remove it from the work queue (LREM).
pub async fn ack(&self, work_key: &str, task_json: &str) -> Result<(), GitError> {
self.ack_raw(work_key, task_json).await
}
async fn ack_raw(&self, work_key: &str, task_json: &str) -> Result<(), GitError> {
let redis = tokio::time::timeout(POOL_GET_TIMEOUT, self.pool.get())
.await
.map_err(|_| GitError::Internal("redis pool get timed out".into()))?
.map_err(|e| GitError::Internal(format!("redis pool get failed: {}", e)))?;
let mut conn: RedisConn = redis;
let _: i64 = redis::cmd("LREM")
.arg(work_key)
.arg(-1)
.arg(task_json)
.query_async(&mut conn)
.await
.map_err(|e| GitError::Internal(format!("LREM failed: {}", e)))?;
Ok(())
}
/// Negative acknowledge (retry): remove from work queue and push back to main queue.
pub async fn nak(
&self,
work_key: &str,
queue_key: &str,
task_json: &str,
) -> Result<(), GitError> {
self.nak_with_retry(work_key, queue_key, task_json, task_json).await
}
/// Negative acknowledge with a different (updated) task JSON — used to
/// requeue with an incremented retry_count.
/// Uses a Lua script for atomic LREM + LPUSH to prevent task loss on crash.
pub async fn nak_with_retry(
&self,
work_key: &str,
queue_key: &str,
old_task_json: &str,
new_task_json: &str,
) -> Result<(), GitError> {
let redis = tokio::time::timeout(POOL_GET_TIMEOUT, self.pool.get())
.await
.map_err(|_| GitError::Internal("redis pool get timed out".into()))?
.map_err(|e| GitError::Internal(format!("redis pool get failed: {}", e)))?;
let mut conn: RedisConn = redis;
// Atomic: remove from work queue AND push to retry queue in one script.
// If the process crashes mid-script, either both happen or neither — no lost tasks.
let script = r#"
redis.call("LREM", KEYS[1], 1, ARGV[1])
redis.call("LPUSH", KEYS[2], ARGV[2])
return 1
"#;
let _: i32 = redis::Script::new(script)
.key(work_key)
.key(queue_key)
.arg(old_task_json)
.arg(new_task_json)
.invoke_async(&mut conn)
.await
.map_err(|e| GitError::Internal(format!("nak script failed: {}", e)))?;
slog::warn!(self.logger, "task nack'd and requeued queue={}", queue_key);
Ok(())
}
pub fn prefix(&self) -> &str {
&self.prefix
}
}
impl Clone for RedisConsumer {
fn clone(&self) -> Self {
Self {
pool: self.pool.clone(),
prefix: self.prefix.clone(),
block_timeout_secs: self.block_timeout_secs,
logger: self.logger.clone(),
}
}
}