add distinguished "Cancelled" error for walredo (don't use it yet)

pageserver/src/tenant.rs

@@ -88,6 +88,7 @@ use crate::tenant::remote_timeline_client::MaybeDeletedIndexPart;
 use crate::tenant::remote_timeline_client::INITDB_PATH;
 use crate::tenant::storage_layer::DeltaLayer;
 use crate::tenant::storage_layer::ImageLayer;
+use crate::walredo::Error;
 use crate::InitializationOrder;
 use std::collections::hash_map::Entry;
 use std::collections::BTreeSet;
@@ -343,7 +344,7 @@ impl WalRedoManager {
         base_img: Option<(Lsn, bytes::Bytes)>,
         records: Vec<(Lsn, crate::walrecord::NeonWalRecord)>,
         pg_version: u32,
-    ) -> anyhow::Result<bytes::Bytes> {
+    ) -> Result<bytes::Bytes, Error> {
         match self {
             Self::Prod(mgr) => {
                 mgr.request_redo(key, lsn, base_img, records, pg_version)
@@ -3953,7 +3954,7 @@ pub(crate) mod harness {
             base_img: Option<(Lsn, Bytes)>,
             records: Vec<(Lsn, NeonWalRecord)>,
             _pg_version: u32,
-        ) -> anyhow::Result<Bytes> {
+        ) -> Result<Bytes, Error> {
             let records_neon = records.iter().all(|r| apply_neon::can_apply_in_neon(&r.1));
             if records_neon {
                 // For Neon wal records, we can decode without spawning postgres, so do so.

pageserver/src/tenant/timeline.rs

@@ -5235,10 +5235,14 @@ impl Timeline {
                     .map_err(PageReconstructError::WalRedo)?
                     .request_redo(key, request_lsn, data.img, data.records, self.pg_version)
                     .await
-                    .context("reconstruct a page image")
                 {
                     Ok(img) => img,
-                    Err(e) => return Err(PageReconstructError::WalRedo(e)),
+                    Err(e) => {
+                        return Err(match e {
+                            crate::walredo::Error::Cancelled => PageReconstructError::Cancelled,
+                            crate::walredo::Error::Other(e) => PageReconstructError::WalRedo(e),
+                        })
+                    }
                 };
 
                 Ok(img)

pageserver/src/walredo.rs

@@ -68,6 +68,20 @@ pub struct PostgresRedoManager {
     redo_process: heavier_once_cell::OnceCell<Arc<process::WalRedoProcess>>,
 }
 
+#[derive(Debug, thiserror::Error)]
+pub enum Error {
+    #[error("cancelled")]
+    Cancelled,
+    #[error(transparent)]
+    Other(#[from] anyhow::Error),
+}
+
+macro_rules! bail {
+    ($($arg:tt)*) => {
+        return Err(Error::Other(anyhow::anyhow!($($arg)*)));
+    }
+}
+
 ///
 /// Public interface of WAL redo manager
 ///
@@ -88,9 +102,9 @@ impl PostgresRedoManager {
         base_img: Option<(Lsn, Bytes)>,
         records: Vec<(Lsn, NeonWalRecord)>,
         pg_version: u32,
-    ) -> anyhow::Result<Bytes> {
+    ) -> Result<Bytes, Error> {
         if records.is_empty() {
-            anyhow::bail!("invalid WAL redo request with no records");
+            bail!("invalid WAL redo request with no records");
         }
 
         let base_img_lsn = base_img.as_ref().map(|p| p.0).unwrap_or(Lsn::INVALID);
@@ -203,7 +217,7 @@ impl PostgresRedoManager {
         records: &[(Lsn, NeonWalRecord)],
         wal_redo_timeout: Duration,
         pg_version: u32,
-    ) -> anyhow::Result<Bytes> {
+    ) -> Result<Bytes, Error> {
         *(self.last_redo_at.lock().unwrap()) = Some(Instant::now());
 
         let (rel, blknum) = key.to_rel_block().context("invalid record")?;
@@ -315,7 +329,7 @@ impl PostgresRedoManager {
             }
             n_attempts += 1;
             if n_attempts > MAX_RETRY_ATTEMPTS || result.is_ok() {
-                return result;
+                return result.map_err(Error::Other);
             }
         }
     }
@@ -329,7 +343,7 @@ impl PostgresRedoManager {
         lsn: Lsn,
         base_img: Option<Bytes>,
         records: &[(Lsn, NeonWalRecord)],
-    ) -> anyhow::Result<Bytes> {
+    ) -> Result<Bytes, Error> {
         let start_time = Instant::now();
 
         let mut page = BytesMut::new();
@@ -338,7 +352,7 @@ impl PostgresRedoManager {
             page.extend_from_slice(&fpi[..]);
         } else {
             // All the current WAL record types that we can handle require a base image.
-            anyhow::bail!("invalid neon WAL redo request with no base image");
+            bail!("invalid neon WAL redo request with no base image");
         }
 
         // Apply all the WAL records in the batch