c/partition_balancer: planner cancelations

planner cancel movement to unavailable nodes
redpanda-data · Jul 6, 2022 · 974815d · 974815d
1 parent 6148d64
commit 974815d
Show file tree

Hide file tree

Showing 3 changed files with 51 additions and 19 deletions.
diff --git a/src/v/cluster/partition_balancer_planner.cc b/src/v/cluster/partition_balancer_planner.cc
@@ -395,26 +395,49 @@ void partition_balancer_planner::get_full_node_reassignments(
     }
 }
 
-std::vector<ntp_reassignments>
-partition_balancer_planner::get_ntp_reassignments(
+void partition_balancer_planner::get_unavailable_node_movement_cancelations(
+  std::vector<model::ntp>& cancelations,
+  const reallocation_request_state& rrs) {
+    for (auto& update : _topic_table.in_progress_updates()) {
+        if (
+          update.second.state
+          != topic_table::in_progress_state::update_requested) {
+            continue;
+        }
+        auto current_assignments = _topic_table.get_partition_assignment(
+          update.first);
+        if (!current_assignments.has_value()) {
+            continue;
+        }
+        for (auto& r : current_assignments->replicas) {
+            if (rrs.unavailable_nodes.contains(r.node_id)) {
+                cancelations.push_back(update.first);
+                continue;
+            }
+        }
+    }
+}
+
+planner_result partition_balancer_planner::get_ntp_reassignments(
   const cluster_health_report& health_report,
   const std::vector<raft::follower_metrics>& follower_metrics) {
+    reallocation_request_state rrs;
+    planner_result result;
+
+    calculate_unavailable_nodes(follower_metrics, rrs);
+
     if (_topic_table.has_updates_in_progress()) {
-        return {};
+        get_unavailable_node_movement_cancelations(result.cancelations, rrs);
+        return result;
     }
 
-    reallocation_request_state rrs;
-    std::vector<ntp_reassignments> reassignments;
-
     init_ntp_sizes_and_node_disk_reports_from_health_report(health_report, rrs);
-
-    calculate_unavailable_nodes(follower_metrics, rrs);
     calculate_nodes_with_disk_constraints_violation(rrs);
 
-    get_unavailable_nodes_reassignments(reassignments, rrs);
-    get_full_node_reassignments(reassignments, rrs);
+    get_unavailable_nodes_reassignments(result.reassignments, rrs);
+    get_full_node_reassignments(result.reassignments, rrs);
 
-    return reassignments;
+    return result;
 }
 
-} // namespace cluster
+} // namespace cluster
diff --git a/src/v/cluster/partition_balancer_planner.h b/src/v/cluster/partition_balancer_planner.h
@@ -19,11 +19,6 @@
 
 namespace cluster {
 
-struct ntp_reassignments {
-    model::ntp ntp;
-    allocation_units allocation_units;
-};
-
 struct planner_config {
     double max_disk_usage_ratio;
     // Size of partitions that can be planned to move in one request
@@ -38,7 +33,7 @@ class partition_balancer_planner {
       topic_table& topic_table,
       partition_allocator& partition_allocator);
 
-    std::vector<ntp_reassignments> get_ntp_reassignments(
+    planner_result get_ntp_reassignments(
       const cluster_health_report&, const std::vector<raft::follower_metrics>&);
 
 private:
@@ -72,6 +67,9 @@ class partition_balancer_planner {
     void get_full_node_reassignments(
       std::vector<ntp_reassignments>&, reallocation_request_state&);
 
+    void get_unavailable_node_movement_cancelations(
+      std::vector<model::ntp>& cancelations, const reallocation_request_state&);
+
     void calculate_nodes_with_disk_constraints_violation(
       reallocation_request_state&);
 

diff --git a/src/v/cluster/partition_balancer_types.h b/src/v/cluster/partition_balancer_types.h
@@ -10,6 +10,7 @@
 
 #pragma once
 
+#include "cluster/scheduling/types.h"
 #include "model/fundamental.h"
 #include "model/metadata.h"
 
@@ -37,4 +38,14 @@ struct node_disk_space {
     }
 };
 
-} // namespace cluster
+struct ntp_reassignments {
+    model::ntp ntp;
+    allocation_units allocation_units;
+};
+
+struct planner_result {
+    std::vector<ntp_reassignments> reassignments;
+    std::vector<model::ntp> cancelations;
+};
+
+} // namespace cluster