Merge pull request 'Store data blocks only on nodes in the latest cluster layout version (fix #815)' (#956) from fix-815 into main

Reviewed-on: https://git.deuxfleurs.fr/Deuxfleurs/garage/pulls/956

2 files changed, 25 insertions, 3 deletions

diff --git a/src/block/manager.rs b/src/block/manager.rs
index 537e1fc1..572bdadd 100644
--- a/src/block/manager.rs
+++ b/src/block/manager.rs
@@ -370,7 +370,7 @@ impl BlockManager {
 		prevent_compression: bool,
 		order_tag: Option<OrderTag>,
 	) -> Result<(), Error> {
-		let who = self.replication.write_sets(&hash);
+		let who = self.system.cluster_layout().current_storage_nodes_of(&hash);
 
 		let compression_level = self.compression_level.filter(|_| !prevent_compression);
 		let (header, bytes) = DataBlock::from_buffer(data, compression_level)
@@ -396,7 +396,7 @@ impl BlockManager {
 			.rpc_helper()
 			.try_write_many_sets(
 				&self.endpoint,
-				who.as_ref(),
+				&[who],
 				put_block_rpc,
 				RequestStrategy::with_priority(PRIO_NORMAL | PRIO_SECONDARY)
 					.with_drop_on_completion(permit)
diff --git a/src/block/resync.rs b/src/block/resync.rs
index 947c68de..b476a0b8 100644
--- a/src/block/resync.rs
+++ b/src/block/resync.rs
@@ -377,7 +377,10 @@ impl BlockResyncManager {
 			info!("Resync block {:?}: offloading and deleting", hash);
 			let existing_path = existing_path.unwrap();
 
-			let mut who = manager.replication.storage_nodes(hash);
+			let mut who = manager
+				.system
+				.cluster_layout()
+				.current_storage_nodes_of(hash);
 			if who.len() < manager.replication.write_quorum() {
 				return Err(Error::Message("Not trying to offload block because we don't have a quorum of nodes to write to".to_string()));
 			}
@@ -455,6 +458,25 @@ impl BlockResyncManager {
 		}
 
 		if rc.is_nonzero() && !exists {
+			// The refcount is > 0, and the block is not present locally.
+			// We might need to fetch it from another node.
+
+			// First, check whether we are still supposed to store that
+			// block in the latest cluster layout version.
+			let storage_nodes = manager
+				.system
+				.cluster_layout()
+				.current_storage_nodes_of(&hash);
+
+			if !storage_nodes.contains(&manager.system.id) {
+				info!(
+					"Resync block {:?}: block is absent with refcount > 0, but it will drop to zero after all metadata is synced. Not fetching the block.",
+					hash
+				);
+				return Ok(());
+			}
+
+			// We know we need the block. Fetch it.
 			info!(
 				"Resync block {:?}: fetching absent but needed block (refcount > 0)",
 				hash