Gap Sync: Skip Body Requests for Non-Archive Nodes

cumulus/zombienet/zombienet-sdk/tests/zombie_ci/full_node_warp_sync/common.rs

@@ -109,6 +109,7 @@ pub(crate) async fn build_network_config(
 						)
 							.into(),
 						("--sync", "warp").into(),
+						("--blocks-pruning", "256").into(),
 					])
 				})
 				.with_node(|node| {

cumulus/zombienet/zombienet-sdk/tests/zombie_ci/full_node_warp_sync/full_node_warp_sync.rs

@@ -86,9 +86,9 @@ async fn assert_warp_sync(node: &NetworkNode) -> Result<(), anyhow::Error> {
 	Ok(())
 }
 
-// Asserting Gap sync requires at least sync=debug level
-async fn assert_gap_sync(node: &NetworkNode) -> Result<(), anyhow::Error> {
-	let option_1_line = LogLineCountOptions::new(|n| n == 1, Duration::from_secs(20), false);
+// Asserting Gap sync requires at least sync=trace level
+async fn assert_gap_sync(node: &NetworkNode, is_archive: bool) -> Result<(), anyhow::Error> {
+	let option_1_line = LogLineCountOptions::new(|n| n == 1, Duration::from_secs(5), false);
 	let option_at_least_5_lines =
 		LogLineCountOptions::new(|n| n >= 5, Duration::from_secs(20), false);
 
@@ -117,6 +117,27 @@ async fn assert_gap_sync(node: &NetworkNode) -> Result<(), anyhow::Error> {
 		return Err(anyhow!("Gap sync block imports are not started"));
 	}
 
+	// Verify body download behavior based on archive mode:
+	// - Archive nodes should download bodies (body: non-zero)
+	// - Non-archive nodes should skip bodies (body: 0 B)
+	let (body_pattern, body_error) = if is_archive {
+		(
+			r"(?<!\[Parachain\] )Gap sync cumulative stats:.*body: [1-9]",
+			"archive node should download bodies",
+		)
+	} else {
+		(
+			r"(?<!\[Parachain\] )Gap sync cumulative stats:.*body: 0 B",
+			"non-archive node should not download bodies",
+		)
+	};
+	let result = node
+		.wait_log_line_count_with_timeout(body_pattern, false, option_at_least_5_lines.clone())
+		.await?;
+	if !result.success() {
+		return Err(anyhow!("Gap sync: {body_error}"));
+	}
+
 	let result = node
 		.wait_log_line_count_with_timeout(
 			r"(?<!\[Parachain\] )Imported [0-9]+ out of [0-9]+ blocks.*\(origin: GapSync\)",
@@ -160,9 +181,10 @@ async fn full_node_warp_sync() -> Result<(), anyhow::Error> {
 
 	// Assert warp and gap syncs only for relaychain.
 	// "five" is not warp syncing the relaychain
-	for name in ["dave", "eve", "four"] {
+	// dave is non-archive (--blocks-pruning 256), eve and four use default (archive-canonical)
+	for (name, is_archive) in [("dave", false), ("eve", true), ("four", true)] {
 		assert_warp_sync(network.get_node(name)?).await?;
-		assert_gap_sync(network.get_node(name)?).await?;
+		assert_gap_sync(network.get_node(name)?, is_archive).await?;
 	}
 
 	// Check relaychain progress
@@ -204,9 +226,9 @@ async fn full_node_warp_sync() -> Result<(), anyhow::Error> {
 	log::info!("Waiting for ferdie to be up");
 	network.get_node("ferdie")?.wait_until_is_up(60u64).await?;
 
-	// Assert warp and gap sync for ferdie
+	// Assert warp and gap sync for ferdie (uses default archive-canonical)
 	assert_warp_sync(network.get_node("ferdie")?).await?;
-	assert_gap_sync(network.get_node("ferdie")?).await?;
+	assert_gap_sync(network.get_node("ferdie")?, true).await?;
 
 	// Check progress for ferdie
 	log::info!("Checking full node ferdie  is syncing");

prdoc/pr_10752.prdoc

@@ -0,0 +1,20 @@
+title: '  Gap Sync: Skip Body Requests for Non-Archive Nodes'
+doc:
+- audience: Node Operator
+  description: |-
+    ### Summary
+    This PR optimizes gap sync bandwidth usage by skipping body requests for non-archive nodes. Bodies are unnecessary during gap sync when the node doesn't maintain full block history, while archive nodes continue to request bodies to preserve complete history.
+    It reduces bandwidth consumption and database size significantly for typical validator/full nodes.
+
+    Additionally added some gap sync statistics for observability:
+    - Introduced `GapSyncStats` to track bandwidth usage: header bytes, body bytes, justification bytes
+    - Logged on gap sync completion to provide visibility into bandwidth savings
+crates:
+- name: sc-network-sync
+  bump: major
+- name: sc-service
+  bump: major
+- name: sc-client-db
+  bump: patch
+- name: sc-cli
+  bump: patch

substrate/client/cli/src/arg_enums.rs

@@ -294,6 +294,9 @@ pub enum SyncMode {
 	/// Download blocks without executing them. Download latest state without proofs.
 	FastUnsafe,
 	/// Prove finality and download the latest state.
+	/// After warp sync completes, the node will have block headers but not bodies for historical
+	/// blocks (unless `blocks-pruning` is set to archive mode). This saves bandwidth while still
+	/// allowing the node to serve as a warp sync source for other nodes.
 	Warp,
 }
 

substrate/client/db/src/lib.rs

@@ -1726,6 +1726,10 @@ impl<Block: BlockT> Backend<Block> {
 			}
 
 			let should_check_block_gap = !header_exists_in_db || !body_exists_in_db;
+			debug!(
+				target: "db",
+				"should_check_block_gap = {should_check_block_gap}",
+			);
 
 			if should_check_block_gap {
 				let update_gap =