c/health_monitor: changed the partition_status chunk size to 8

With large number of topics that contain a small amount of partitions we
may end up wasting a lot of memory using the default `ss::chunked_fifo`
chunk size of 128.
Change the chunk size of `partition_status` `chunked_fifo` collection to
8 to minimize a waste with large number of small topics.

Signed-off-by: Michal Maslanka <michal@redpanda.com>
(cherry picked from commit 75c9549)

src/v/cluster/health_monitor_backend.cc

@@ -735,8 +735,8 @@ ss::chunked_fifo<ntp_report> collect_shard_local_reports(
     return reports;
 }
 
-using reports_acc_t = absl::
-  node_hash_map<model::topic_namespace, ss::chunked_fifo<partition_status>>;
+using reports_acc_t
+  = absl::node_hash_map<model::topic_namespace, partition_statuses_t>;
 
 reports_acc_t reduce_reports_map(
   reports_acc_t acc, ss::chunked_fifo<ntp_report> current_reports) {

src/v/cluster/health_monitor_types.cc

@@ -14,6 +14,7 @@
 #include "cluster/errc.h"
 #include "cluster/node/types.h"
 #include "features/feature_table.h"
+#include "health_monitor_types.h"
 #include "model/adl_serde.h"
 #include "model/metadata.h"
 #include "utils/to_string.h"
@@ -156,7 +157,7 @@ topic_status& topic_status::operator=(const topic_status& rhs) {
         return *this;
     }
 
-    ss::chunked_fifo<partition_status> p;
+    partition_statuses_t p;
     p.reserve(rhs.partitions.size());
     std::copy(
       rhs.partitions.begin(), rhs.partitions.end(), std::back_inserter(p));
@@ -167,7 +168,7 @@ topic_status& topic_status::operator=(const topic_status& rhs) {
 }
 
 topic_status::topic_status(
-  model::topic_namespace tp_ns, ss::chunked_fifo<partition_status> partitions)
+  model::topic_namespace tp_ns, partition_statuses_t partitions)
   : tp_ns(std::move(tp_ns))
   , partitions(std::move(partitions)) {}
 

src/v/cluster/health_monitor_types.h

@@ -97,21 +97,26 @@ struct partition_status
     friend bool operator==(const partition_status&, const partition_status&)
       = default;
 };
+/**
+ * We keep the chunk size in partition status small to make sure that we do not
+ * waste to much of memory for topics with small number of partitions.
+ */
+using partition_statuses_t = ss::chunked_fifo<partition_status, 8>;
 
 struct topic_status
   : serde::envelope<topic_status, serde::version<0>, serde::compat_version<0>> {
     static constexpr int8_t current_version = 0;
 
     topic_status() = default;
-    topic_status(model::topic_namespace, ss::chunked_fifo<partition_status>);
+    topic_status(model::topic_namespace, partition_statuses_t);
     topic_status& operator=(const topic_status&);
     topic_status(const topic_status&);
     topic_status& operator=(topic_status&&) = default;
     topic_status(topic_status&&) = default;
     ~topic_status() = default;
 
     model::topic_namespace tp_ns;
-    ss::chunked_fifo<partition_status> partitions;
+    partition_statuses_t partitions;
     friend std::ostream& operator<<(std::ostream&, const topic_status&);
     friend bool operator==(const topic_status&, const topic_status&);
 

src/v/cluster/tests/health_monitor_test.cc

@@ -378,7 +378,7 @@ enum part_status { HEALTHY, LEADERLESS, URP };
 
 topic_status
 make_ts(ss::sstring name, const std::vector<part_status>& status_list) {
-    ss::chunked_fifo<cluster::partition_status> statuses;
+    cluster::partition_statuses_t statuses;
 
     partition_id pid{0};
     for (auto status : status_list) {

src/v/cluster/tests/partition_balancer_simulator_test.cc

@@ -609,7 +609,7 @@ class partition_balancer_sim_fixture {
 
             absl::flat_hash_map<
               model::topic_namespace,
-              ss::chunked_fifo<cluster::partition_status>>
+              cluster::partition_statuses_t>
               topic2partitions;
             for (const auto& [ntp, repl] : replicas) {
                 topic2partitions[model::topic_namespace(ntp.ns, ntp.tp.topic)]

src/v/cluster/tests/randoms.h

@@ -18,6 +18,8 @@
 #include "storage/tests/randoms.h"
 #include "test_utils/randoms.h"
 
+#include <iterator>
+
 namespace cluster::node {
 
 inline local_state random_local_state() {
@@ -60,9 +62,10 @@ inline partition_status random_partition_status() {
 }
 
 inline topic_status random_topic_status() {
-    return topic_status(
-      model::random_topic_namespace(),
-      tests::random_chunked_fifo(random_partition_status));
+    auto d = tests::random_vector(random_partition_status);
+    cluster::partition_statuses_t partitions;
+    std::move(d.begin(), d.end(), std::back_inserter(partitions));
+    return {model::random_topic_namespace(), std::move(partitions)};
 }
 
 inline drain_manager::drain_status random_drain_status() {

src/v/cluster/tests/serialization_rt_test.cc

@@ -731,7 +731,7 @@ cluster::drain_manager::drain_status random_drain_status() {
 }
 
 cluster::topic_status random_topic_status() {
-    ss::chunked_fifo<cluster::partition_status> partitions;
+    cluster::partition_statuses_t partitions;
     for (int i = 0, mi = random_generators::get_int(10); i < mi; i++) {
         partitions.push_back(cluster::partition_status{
           .id = tests::random_named_int<model::partition_id>(),

src/v/compat/cluster_json.h

@@ -398,7 +398,7 @@ inline void read_value(json::Value const& rd, cluster::partition_status& obj) {
 
 inline void read_value(json::Value const& rd, cluster::topic_status& obj) {
     model::topic_namespace tp_ns;
-    ss::chunked_fifo<cluster::partition_status> partitions;
+    cluster::partition_statuses_t partitions;
 
     read_member(rd, "tp_ns", tp_ns);
     read_member(rd, "partitions", partitions);