From 22368b13f2b6c7abcbc04431dbb26aa25ec13771 Mon Sep 17 00:00:00 2001
From: Gleb Natapov' via ScyllaDB development <scylladb-dev@googlegroups.com>
Date: Sun, 29 Sep 2024 11:21:38 +0300
Subject: [PATCH] api: introduce raft stepdown REST API

Also provide test.py util function to trigger it. Can be useful for
testing.
---
 api/api-doc/raft.json | 32 ++++++++++++++++++++++++++++++++
 api/raft.cc           | 36 ++++++++++++++++++++++++++++++++++++
 test/topology/util.py |  5 +++++
 3 files changed, 73 insertions(+)

diff --git a/api/api-doc/raft.json b/api/api-doc/raft.json
index 971475eb9a..77e8698b49 100644
--- a/api/api-doc/raft.json
+++ b/api/api-doc/raft.json
@@ -94,6 +94,38 @@
                ]
             }
          ]
+      },
+      {
+         "path":"/raft/trigger_stepdown/",
+         "operations":[
+            {
+               "method":"POST",
+               "summary":"Triggers stepdown of a leader for given Raft group or group0 if not provided (returns an error if the node is not a leader)",
+               "type":"string",
+               "nickname":"trigger_stepdown",
+               "produces":[
+                  "application/json"
+               ],
+               "parameters":[
+                  {
+                     "name":"group_id",
+                     "description":"The ID of the group which leader should stepdown",
+                     "required":false,
+                     "allowMultiple":false,
+                     "type":"string",
+                     "paramType":"query"
+                  },
+                  {
+                     "name":"timeout",
+                     "description":"Timeout in seconds after which the endpoint returns a failure. If not provided, 60s is used.",
+                     "required":false,
+                     "allowMultiple":false,
+                     "type":"long",
+                     "paramType":"query"
+                  }
+               ]
+            }
+         ]
       }
    ]
 }
diff --git a/api/raft.cc b/api/raft.cc
index bdf83b2358..1dcd1254fc 100644
--- a/api/raft.cc
+++ b/api/raft.cc
@@ -11,6 +11,7 @@
 #include "api/api-doc/raft.json.hh"
 
 #include "service/raft/raft_group_registry.hh"
+#include "service/raft/raft_address_map.hh"
 #include "log.hh"
 
 using namespace seastar::httpd;
@@ -123,6 +124,40 @@ void set_raft(http_context&, httpd::routes& r, sharded<service::raft_group_regis
             throw bad_param_exception{fmt::format("Server for group ID {} not found", gid)};
         }
 
+        co_return json_void{};
+    });
+    r::trigger_stepdown.set(r, [&raft_gr] (std::unique_ptr<http::request> req) -> future<json_return_type> {
+        auto timeout = get_request_timeout(*req);
+        auto dur = timeout.value ? *timeout.value - lowres_clock::now() : std::chrono::seconds(60);
+        const auto stepdown_timeout_ticks = dur / service::raft_tick_interval;
+        auto timeout_dur = raft::logical_clock::duration(stepdown_timeout_ticks);
+
+        if (!req->query_parameters.contains("group_id")) {
+            // Stepdown on group 0 by default
+            co_await raft_gr.invoke_on(0, [timeout_dur] (service::raft_group_registry& raft_gr) {
+                apilog.info("Triggering stepdown for group0");
+                return raft_gr.group0().stepdown(timeout_dur);
+            });
+            co_return json_void{};
+        }
+        raft::group_id gid{utils::UUID{req->get_path_param("group_id")}};
+
+        std::atomic<bool> found_srv{false};
+        co_await raft_gr.invoke_on_all([gid, timeout_dur, &found_srv] (service::raft_group_registry& raft_gr) -> future<> {
+            auto* srv = raft_gr.find_server(gid);
+            if (!srv) {
+                co_return;
+            }
+
+            found_srv = true;
+            apilog.info("Triggering stepdown for group {}", gid);
+            co_await srv->stepdown(timeout_dur);
+        });
+
+        if (!found_srv) {
+            throw std::runtime_error{fmt::format("Server for group ID {} not found", gid)};
+        }
+
         co_return json_void{};
     });
 }
@@ -131,6 +166,7 @@ void unset_raft(http_context&, httpd::routes& r) {
     r::trigger_snapshot.unset(r);
     r::get_leader_host.unset(r);
     r::read_barrier.unset(r);
+    r::trigger_stepdown.unset(r);
 }
 
 }
diff --git a/test/topology/util.py b/test/topology/util.py
index afa3c7acd2..5cdac13ed8 100644
--- a/test/topology/util.py
+++ b/test/topology/util.py
@@ -398,6 +398,11 @@ async def trigger_snapshot(manager, server: ServerInfo) -> None:
     host = cql.cluster.metadata.get_host(server.ip_addr)
     await manager.api.client.post(f"/raft/trigger_snapshot/{group0_id}", host=server.ip_addr)
 
+async def trigger_stepdown(manager, server: ServerInfo) -> None:
+    cql = manager.get_cql()
+    host = cql.cluster.metadata.get_host(server.ip_addr)
+    await manager.api.client.post("/raft/trigger_stepdown", host=server.ip_addr)
+
 
 
 async def get_coordinator_host_ids(manager: ManagerClient) -> list[str]: