Allow configuring the max number of chunk retries

Before, this was hard-coded to 10.

Author: Qqwy

opsqueue/src/common/chunk.rs

@@ -367,11 +367,11 @@ pub mod db {
         chunk_id: ChunkId,
         failure: String,
         mut conn: impl WriterConnection,
+        max_retries: u32,
     ) -> sqlx::Result<bool> {
         let failed_permanently = conn
             .transaction(move |mut tx| {
                 Box::pin(async move {
-                    const MAX_RETRIES: i64 = 10;
                     let ChunkId {
                         submission_id,
                         chunk_index,
@@ -388,7 +388,7 @@ pub mod db {
                     .fetch_one(tx.get_inner())
                     .await?;
                     tracing::trace!("Retries: {}", fields.retries);
-                    if fields.retries >= MAX_RETRIES {
+                    if fields.retries >= max_retries.into() {
                         crate::common::submission::db::fail_submission_notx(
                             submission_id,
                             chunk_index,

opsqueue/src/config.rs

@@ -69,6 +69,11 @@ pub struct Config {
     /// At that time, the connection will be closed and any open reservations will be canceled.
     #[arg(long, default_value_t = 3)]
     pub max_missable_heartbeats: usize,
+
+    /// Maximum number of times a chunk is retried before permanently failiing
+    /// the full submission the chunk is a part of.
+    #[arg(long, default_value_t = 10)]
+    pub max_chunk_retries: u32,
 }
 
 impl Default for Config {

opsqueue/src/consumer/server/mod.rs

@@ -72,7 +72,8 @@ impl ServerState {
         config: &'static Config,
     ) -> Self {
         let dispatcher = Dispatcher::new(reservation_expiration);
-        let (completer, completer_tx) = Completer::new(pool.writer_pool(), &dispatcher);
+        let (completer, completer_tx) =
+            Completer::new(pool.writer_pool(), &dispatcher, config.max_chunk_retries);
         Self {
             pool,
             completer: Some(completer),
@@ -170,12 +171,14 @@ pub struct Completer {
     pool: db::WriterPool,
     dispatcher: Dispatcher,
     count: usize,
+    max_chunk_retries: u32,
 }
 
 impl Completer {
     pub fn new(
         pool: &db::WriterPool,
         dispatcher: &Dispatcher,
+        max_chunk_retries: u32,
     ) -> (Self, tokio::sync::mpsc::Sender<CompleterMessage>) {
         let (tx, rx) = tokio::sync::mpsc::channel(1024);
         let pool = pool.clone();
@@ -184,6 +187,7 @@ impl Completer {
             pool: pool.clone(),
             dispatcher: dispatcher.clone(),
             count: 0,
+            max_chunk_retries,
         };
         (me, tx)
     }
@@ -243,8 +247,13 @@ impl Completer {
                 } => {
                     // Even in the unlikely event that the DB write fails,
                     // we still want to unreserve the chunk
-                    let failed_permanently =
-                        crate::common::chunk::db::retry_or_fail_chunk(id, failure, &mut conn).await;
+                    let failed_permanently = crate::common::chunk::db::retry_or_fail_chunk(
+                        id,
+                        failure,
+                        &mut conn,
+                        self.max_chunk_retries,
+                    )
+                    .await;
                     reservations.lock().expect("No poison").remove(&id);
                     let maybe_started_at = self
                         .dispatcher