chore(test): more complex setting for ring test

Signed-off-by: Wataru Ishida <wataru.ishid@gmail.com>
osrg · Mar 12, 2024 · 5d25487 · 5d25487
1 parent f6e89a9
commit 5d25487
Show file tree

Hide file tree

Showing 2 changed files with 5 additions and 2 deletions.
diff --git a/Dockerfile b/Dockerfile
@@ -47,6 +47,7 @@ RUN cd reduction_server && cargo build -r
 FROM optcast AS unittest
 
 ENV RUST_LOG=info
+ENV NCCL_SOCKET_IFNAME=lo
 RUN cd reduction_server && cargo test --all -- --nocapture
 
 FROM nvcr.io/nvidia/cuda:12.3.1-devel-ubuntu22.04 AS final

diff --git a/reduction_server/src/ring.rs b/reduction_server/src/ring.rs
@@ -603,15 +603,17 @@ mod tests {
             .map(|i| {
                 std::thread::spawn(move || {
                     let ring_rank = format!("{}", i + 1);
-                    let address = format!("127.0.0.1:{}", 9090 + i);
+                    let address = format!("127.0.0.1:{},127.0.0.1:{}", 9090 + i, 9100 + i);
                     let args = Args::parse_from([
                         "--bench",
                         "--nrank",
                         "4",
                         "--reduce-threads",
-                        "1",
+                        "2",
                         "--address",
                         &address,
+                        "--nreq",
+                        "1", // when using socket plugin, concurrent recv/send requests doesn't work
                         "--ring-rank",
                         &ring_rank,
                     ]);