[release/6.x] Partial revert of #7314, use node's local snapshot when available (#7589)

Author: eddyashton

CHANGELOG.md

@@ -5,6 +5,14 @@ All notable changes to this project will be documented in this file.
 The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/)
 and this project adheres to [Semantic Versioning](http://semver.org/spec/v2.0.0.html).
 
+## [6.0.19]
+
+[6.0.19]: https://github.com/microsoft/CCF/releases/tag/6.0.19
+
+### Changed
+
+- Partial revert to `fetch_recent_snapshot` behaviour. Nodes will now use a local snapshot if available and sufficiently fresh, and only prefer the peer's snapshot if it is newer (#7589).
+
 ## [6.0.18]
 
 [6.0.18]: https://github.com/microsoft/CCF/releases/tag/6.0.18

python/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "ccf"
-version = "6.0.18"
+version = "6.0.19"
 authors = [
   { name="CCF Team", email="[email protected]" },
 ]

src/host/main.cpp

@@ -886,9 +886,14 @@ int main(int argc, char** argv) // NOLINT(bugprone-exception-escape)
         config.command.join.fetch_recent_snapshot)
       {
         // Try to fetch a recent snapshot from peer
+        const size_t latest_local_idx = latest_local_snapshot.has_value() ?
+          snapshots::get_snapshot_idx_from_file_name(
+            latest_local_snapshot->second) :
+          0;
         auto latest_peer_snapshot = snapshots::fetch_from_peer(
           config.command.join.target_rpc_address,
           config.command.service_certificate_file,
+          latest_local_idx,
           config.command.join.fetch_snapshot_max_attempts,
           config.command.join.fetch_snapshot_retry_interval.count_ms(),
           config.command.join.fetch_snapshot_max_size.count_bytes());
@@ -908,10 +913,9 @@ int main(int argc, char** argv) // NOLINT(bugprone-exception-escape)
             fs::path(latest_peer_snapshot->snapshot_name);
           if (files::exists(dst_path))
           {
-            LOG_FAIL_FMT(
-              "Overwriting existing snapshot at {} with data retrieved from "
-              "peer",
-              dst_path);
+            throw std::logic_error(fmt::format(
+              "Unable to write peer snapshot - already have a file at {}",
+              dst_path));
           }
           files::dump(latest_peer_snapshot->snapshot_data, dst_path);
           startup_snapshot = latest_peer_snapshot->snapshot_data;

src/snapshots/fetch.h

@@ -121,6 +121,7 @@ namespace snapshots
   static std::optional<SnapshotResponse> try_fetch_from_peer(
     const std::string& peer_address,
     const std::string& path_to_peer_ca,
+    size_t latest_local_snapshot,
     size_t max_size)
   {
     try
@@ -135,8 +136,10 @@ namespace snapshots
       // redirects terminate, the final response is likely to be extremely large
       // so is fetched over multiple requests for a sub-range, returning
       // PARTIAL_CONTENT each time.
-      std::string snapshot_url =
-        fmt::format("https://{}/node/snapshot", peer_address);
+      std::string snapshot_url = fmt::format(
+        "https://{}/node/snapshot?since={}",
+        peer_address,
+        latest_local_snapshot);
 
       // Fetch 4MB chunks at a time
       constexpr size_t range_size = 4L * 1024 * 1024;
@@ -239,7 +242,8 @@ namespace snapshots
 
         if (status_code == HTTP_STATUS_NOT_FOUND)
         {
-          LOG_INFO_FMT("Peer has no suitable snapshot");
+          LOG_INFO_FMT(
+            "Peer has no snapshot newer than {}", latest_local_snapshot);
           return std::nullopt;
         }
 
@@ -340,6 +344,7 @@ namespace snapshots
   static std::optional<SnapshotResponse> fetch_from_peer(
     const std::string& peer_address,
     const std::string& path_to_peer_ca,
+    size_t latest_local_snapshot,
     size_t max_attempts,
     size_t retry_delay_ms,
     size_t max_size)
@@ -357,8 +362,8 @@ namespace snapshots
         std::this_thread::sleep_for(std::chrono::milliseconds(retry_delay_ms));
       }
 
-      auto response =
-        try_fetch_from_peer(peer_address, path_to_peer_ca, max_size);
+      auto response = try_fetch_from_peer(
+        peer_address, path_to_peer_ca, latest_local_snapshot, max_size);
       if (response.has_value())
       {
         return response;