Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

kafka: Add fetch plan and execute latency metric #13485

Merged
merged 1 commit into from
Nov 8, 2023
Merged

kafka: Add fetch plan and execute latency metric #13485

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
38 changes: 38 additions & 0 deletions src/v/kafka/latency_probe.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -18,6 +18,8 @@

#include <seastar/core/metrics.hh>

#include <chrono>

namespace kafka {
class latency_probe {
public:
Expand All @@ -36,8 +38,10 @@ class latency_probe {
if (config::shard_local_cfg().disable_metrics()) {
return;
}

std::vector<sm::label_instance> labels{
sm::label("latency_metric")("microseconds")};

_metrics.add_group(
prometheus_sanitize::metrics_name("kafka:latency"),
{
Expand All @@ -54,6 +58,29 @@ class latency_probe {
},
{},
{sm::shard_label});

auto add_plan_and_execute_metric =
[this, &labels](const std::string& fetch_label, hist_t& hist) {
auto fetch_labels = labels;
auto fetch_result_label = sm::label("fetch_result")(fetch_label);
fetch_labels.push_back(fetch_result_label);
_metrics.add_group(
prometheus_sanitize::metrics_name("fetch_stats"),
{// Measures latency off creating the fetch plan and
// subsequently executing it - aka "one poll loop"
sm::make_histogram(
"plan_and_execute_latency_us",
sm::description("Latency of fetch planning and excution"),
fetch_labels,
[&hist] { return hist.internal_histogram_logform(); })},
{},
{sm::shard_label});
};

add_plan_and_execute_metric(
"non-empty", _fetch_plan_and_execute_latency);
add_plan_and_execute_metric(
"empty", _fetch_plan_and_execute_latency_empty);
}

void setup_public_metrics() {
Expand Down Expand Up @@ -88,9 +115,20 @@ class latency_probe {
_fetch_latency.record(micros.count());
}

void record_fetch_plan_and_execute_measurement(
std::chrono::microseconds micros, bool empty) {
if (empty) {
_fetch_plan_and_execute_latency_empty.record(micros.count());
} else {
_fetch_plan_and_execute_latency.record(micros.count());
}
}

private:
hist_t _produce_latency;
hist_t _fetch_latency;
hist_t _fetch_plan_and_execute_latency;
hist_t _fetch_plan_and_execute_latency_empty;
metrics::internal_metric_groups _metrics;
metrics::public_metric_groups _public_metrics;
};
Expand Down
12 changes: 12 additions & 0 deletions src/v/kafka/server/handlers/fetch.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -13,6 +13,7 @@
#include "cluster/partition_manager.h"
#include "cluster/shard_table.h"
#include "config/configuration.h"
#include "kafka/latency_probe.h"
#include "kafka/protocol/batch_consumer.h"
#include "kafka/protocol/errors.h"
#include "kafka/protocol/fetch.h"
Expand Down Expand Up @@ -836,6 +837,9 @@ class simple_fetch_planner final : public fetch_planner::impl {
*/

static ss::future<> fetch_topic_partitions(op_context& octx) {
auto bytes_left_before = octx.bytes_left;

auto start_time = latency_probe::hist_t::clock_type::now();
auto planner = make_fetch_planner<simple_fetch_planner>();

auto fetch_plan = planner.create_plan(octx);
Expand All @@ -844,6 +848,14 @@ static ss::future<> fetch_topic_partitions(op_context& octx) {
= make_fetch_plan_executor<parallel_fetch_plan_executor>();
co_await executor.execute_plan(octx, std::move(fetch_plan));

auto end_time = latency_probe::hist_t::clock_type::now();

auto latency = std::chrono::duration_cast<std::chrono::microseconds>(
end_time - start_time);

octx.rctx.probe().record_fetch_plan_and_execute_measurement(
latency, bytes_left_before == octx.bytes_left);

if (octx.should_stop_fetch()) {
co_return;
}
Expand Down
Loading