ignore maintenance

rystsov · Sep 27, 2023 · c9f3401 · c9f3401
1 parent 0cb37cc
commit c9f3401
Show file tree

Hide file tree

Showing 4 changed files with 14 additions and 27 deletions.
diff --git a/src/v/cluster/partition_balancer_backend.cc b/src/v/cluster/partition_balancer_backend.cc
@@ -381,9 +381,7 @@ ss::future<> partition_balancer_backend::do_tick() {
         _cur_term->last_status = partition_balancer_status::in_progress;
     } else if (plan_data.status == planner_status::waiting_for_reports) {
         _cur_term->last_status = partition_balancer_status::starting;
-    } else if (
-      plan_data.failed_actions_count > 0
-      || plan_data.status == planner_status::waiting_for_maintenance_end) {
+    } else if (plan_data.failed_actions_count > 0) {
         _cur_term->last_status = partition_balancer_status::stalled;
     } else {
         _cur_term->last_status = partition_balancer_status::ready;

diff --git a/src/v/cluster/partition_balancer_planner.cc b/src/v/cluster/partition_balancer_planner.cc
@@ -77,7 +77,6 @@ class partition_balancer_planner::request_context {
     std::vector<model::node_id> all_nodes;
     absl::flat_hash_set<model::node_id> all_unavailable_nodes;
     absl::flat_hash_set<model::node_id> timed_out_unavailable_nodes;
-    size_t num_nodes_in_maintenance = 0;
     absl::flat_hash_set<model::node_id> decommissioning_nodes;
     absl::flat_hash_map<model::node_id, node_disk_space> node_disk_reports;
 
@@ -237,13 +236,6 @@ void partition_balancer_planner::init_per_node_state(
 
         ctx.all_nodes.push_back(id);
 
-        if (
-          broker.state.get_maintenance_state()
-          == model::maintenance_state::active) {
-            vlog(clusterlog.debug, "node {}: in maintenance", id);
-            ctx.num_nodes_in_maintenance += 1;
-        }
-
         if (
           broker.state.get_membership_state()
           == model::membership_state::draining) {
@@ -1598,16 +1590,12 @@ partition_balancer_planner::plan_actions(
       ctx, ctx.decommissioning_nodes, change_reason::node_decommissioning);
 
     if (ctx.config().mode == model::partition_autobalancing_mode::continuous) {
-        if (ctx.num_nodes_in_maintenance == 0) {
-            co_await get_node_drain_actions(
-              ctx,
-              ctx.timed_out_unavailable_nodes,
-              change_reason::node_unavailable);
-            co_await get_rack_constraint_repair_actions(ctx);
-            co_await get_full_node_actions(ctx);
-        } else if (!result.violations.is_empty()) {
-            result.status = status::waiting_for_maintenance_end;
-        }
+        co_await get_node_drain_actions(
+          ctx,
+          ctx.timed_out_unavailable_nodes,
+          change_reason::node_unavailable);
+        co_await get_rack_constraint_repair_actions(ctx);
+        co_await get_full_node_actions(ctx);
     }
     co_await get_counts_rebalancing_actions(ctx);
 

diff --git a/src/v/cluster/partition_balancer_planner.h b/src/v/cluster/partition_balancer_planner.h
@@ -63,7 +63,6 @@ class partition_balancer_planner {
     enum class status {
         empty,
         actions_planned,
-        waiting_for_maintenance_end,
         waiting_for_reports,
         missing_sizes,
     };

diff --git a/src/v/cluster/tests/partition_balancer_planner_test.cc b/src/v/cluster/tests/partition_balancer_planner_test.cc
@@ -743,11 +743,13 @@ FIXTURE_TEST(
     auto plan_data = planner.plan_actions(hr, as).get();
 
     check_violations(plan_data, unavailable_nodes, {});
-    BOOST_REQUIRE_EQUAL(plan_data.reassignments.size(), 0);
-    BOOST_REQUIRE(
-      plan_data.status
-      == cluster::partition_balancer_planner::status::
-        waiting_for_maintenance_end);
+    BOOST_REQUIRE_EQUAL(plan_data.reassignments.size(), 1);
+
+    std::unordered_set<model::node_id> expected_nodes(
+      {model::node_id(1), model::node_id(2), model::node_id(3)});
+
+    auto new_replicas = plan_data.reassignments.front().allocated.replicas();
+    check_expected_assignments(new_replicas, expected_nodes);
 }
 
 /*