[pd-router] Add Configurable Retry Logic for reduce backend pressure (#8744)

2025-08-04 20:42:07 -07:00
parent d98a4913ea
commit 354ac43555
10 changed files with 502 additions and 293 deletions
--- a/sgl-router/src/routers/factory.rs
+++ b/sgl-router/src/routers/factory.rs
@@ -50,6 +50,7 @@ impl RouterFactory {
            ctx.router_config.worker_startup_check_interval_secs,
            ctx.router_config.dp_aware,
            ctx.router_config.api_key.clone(),
+            ctx.router_config.retry.clone(),
        )?;

        Ok(Box::new(router))
@@ -79,6 +80,7 @@ impl RouterFactory {
            ctx.client.clone(),
            ctx.router_config.worker_startup_timeout_secs,
            ctx.router_config.worker_startup_check_interval_secs,
+            ctx.router_config.retry.clone(),
        )?;

        Ok(Box::new(router))