SERVER-120069 Instrument the storage layer with OpenTelemetry Metrics (#53848)
GitOrigin-RevId: 4f19ab5bf8733201b10d1ded3d6988faba5a99c3
This commit is contained in:
DennisSHCheung
MongoDB Bot
parent
005ea9599c
commit
837806ab05
@ -2810,6 +2810,8 @@ mongo_cc_library(
|
||||
"startup_warnings_mongod",
|
||||
"//src/mongo/db/storage:backup_cursor_hooks",
|
||||
"//src/mongo/db/storage:disk_space_monitor",
|
||||
"//src/mongo/otel/metrics/instrumentation:disk_metrics",
|
||||
"//src/mongo/otel/metrics/instrumentation:system_mount_metrics",
|
||||
"//src/mongo/db/storage:oplog_truncation",
|
||||
"//src/mongo/db/storage:storage_control",
|
||||
"system_index",
|
||||
|
||||
@ -230,6 +230,8 @@
|
||||
#include "mongo/executor/task_executor_pool.h"
|
||||
#include "mongo/executor/thread_pool_task_executor.h"
|
||||
#include "mongo/logv2/log.h"
|
||||
#include "mongo/otel/metrics/instrumentation/disk_metrics.h"
|
||||
#include "mongo/otel/metrics/instrumentation/system_mount_metrics.h"
|
||||
#include "mongo/otel/metrics/metrics_initialization.h"
|
||||
#include "mongo/otel/traces/trace_initialization.h"
|
||||
#include "mongo/platform/atomic_word.h"
|
||||
@ -1011,6 +1013,8 @@ ExitCode _initAndListen(ServiceContext* serviceContext) {
|
||||
startFLECrud(serviceContext);
|
||||
|
||||
DiskSpaceMonitor::start(serviceContext);
|
||||
installSystemMountOtelMetrics(serviceContext);
|
||||
installDiskOtelMetrics(serviceContext);
|
||||
if (!storageEngine->storesFilesInDbPath()) {
|
||||
LOGV2(7333400,
|
||||
"The index builds DiskSpaceMonitor action which periodically checks if we "
|
||||
|
||||
@ -381,5 +381,7 @@ test_suite(
|
||||
":otel_metric_name_validation_test",
|
||||
":otel_metric_server_status_adapter_test",
|
||||
":server_status_metric_name_validation_test",
|
||||
"//src/mongo/otel/metrics/instrumentation:disk_metrics_test",
|
||||
"//src/mongo/otel/metrics/instrumentation:system_mount_metrics_test",
|
||||
],
|
||||
)
|
||||
|
||||
71
src/mongo/otel/metrics/instrumentation/BUILD.bazel
Normal file
71
src/mongo/otel/metrics/instrumentation/BUILD.bazel
Normal file
@ -0,0 +1,71 @@
|
||||
load("//bazel:mongo_src_rules.bzl", "mongo_cc_library", "mongo_cc_unit_test")
|
||||
|
||||
package(default_visibility = ["//visibility:public"])
|
||||
|
||||
exports_files(
|
||||
glob([
|
||||
"*.h",
|
||||
]),
|
||||
)
|
||||
|
||||
mongo_cc_library(
|
||||
name = "disk_metrics",
|
||||
srcs_select = [{
|
||||
"@platforms//os:linux": ["disk_metrics_linux.cpp"],
|
||||
"//conditions:default": ["disk_metrics_noop.cpp"],
|
||||
}],
|
||||
deps = [
|
||||
"//src/mongo/db:service_context",
|
||||
] + select({
|
||||
"@platforms//os:linux": [
|
||||
"//src/mongo/otel/metrics:otel_metrics_service",
|
||||
"//src/mongo/util:periodic_runner",
|
||||
"//src/mongo/util:procparser",
|
||||
],
|
||||
"//conditions:default": [],
|
||||
}),
|
||||
)
|
||||
|
||||
mongo_cc_unit_test(
|
||||
name = "disk_metrics_test",
|
||||
srcs = ["disk_metrics_test.cpp"],
|
||||
tags = ["mongo_unittest_fourth_group"],
|
||||
deps = [
|
||||
":disk_metrics",
|
||||
"//src/mongo/db:service_context_test_fixture",
|
||||
"//src/mongo/db/auth:authmocks",
|
||||
"//src/mongo/db/shard_role:service_context_non_d",
|
||||
"//src/mongo/otel/metrics:metrics_test_util",
|
||||
],
|
||||
)
|
||||
|
||||
mongo_cc_library(
|
||||
name = "system_mount_metrics",
|
||||
srcs_select = [{
|
||||
"@platforms//os:linux": ["system_mount_metrics_linux.cpp"],
|
||||
"//conditions:default": ["system_mount_metrics_noop.cpp"],
|
||||
}],
|
||||
deps = [
|
||||
"//src/mongo/db:service_context",
|
||||
] + select({
|
||||
"@platforms//os:linux": [
|
||||
"//src/mongo/otel/metrics:otel_metrics_service",
|
||||
"//src/mongo/util:periodic_runner",
|
||||
"//src/mongo/util:procparser",
|
||||
],
|
||||
"//conditions:default": [],
|
||||
}),
|
||||
)
|
||||
|
||||
mongo_cc_unit_test(
|
||||
name = "system_mount_metrics_test",
|
||||
srcs = ["system_mount_metrics_test.cpp"],
|
||||
tags = ["mongo_unittest_fifth_group"],
|
||||
deps = [
|
||||
":system_mount_metrics",
|
||||
"//src/mongo/db:service_context_test_fixture",
|
||||
"//src/mongo/db/auth:authmocks",
|
||||
"//src/mongo/db/shard_role:service_context_non_d",
|
||||
"//src/mongo/otel/metrics:metrics_test_util",
|
||||
],
|
||||
)
|
||||
72
src/mongo/otel/metrics/instrumentation/disk_metrics.h
Normal file
72
src/mongo/otel/metrics/instrumentation/disk_metrics.h
Normal file
@ -0,0 +1,72 @@
|
||||
/**
|
||||
* Copyright (C) 2026-present MongoDB, Inc.
|
||||
*
|
||||
* This program is free software: you can redistribute it and/or modify
|
||||
* it under the terms of the Server Side Public License, version 1,
|
||||
* as published by MongoDB, Inc.
|
||||
*
|
||||
* This program is distributed in the hope that it will be useful,
|
||||
* but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
||||
* Server Side Public License for more details.
|
||||
*
|
||||
* You should have received a copy of the Server Side Public License
|
||||
* along with this program. If not, see
|
||||
* <http://www.mongodb.com/licensing/server-side-public-license>.
|
||||
*
|
||||
* As a special exception, the copyright holders give permission to link the
|
||||
* code of portions of this program with the OpenSSL library under certain
|
||||
* conditions as described in each individual source file and distribute
|
||||
* linked combinations including the program with the OpenSSL library. You
|
||||
* must comply with the Server Side Public License in all respects for
|
||||
* all of the code used other than as permitted herein. If you modify file(s)
|
||||
* with this exception, you may extend this exception to your version of the
|
||||
* file(s), but you are not obligated to do so. If you do not wish to do so,
|
||||
* delete this exception statement from your version. If you delete this
|
||||
* exception statement from all source files in the program, then also delete
|
||||
* it in the license file.
|
||||
*/
|
||||
|
||||
#pragma once
|
||||
|
||||
#include "mongo/bson/bsonobj.h"
|
||||
#include "mongo/util/modules.h"
|
||||
|
||||
#include <memory>
|
||||
#include <string>
|
||||
#include <vector>
|
||||
|
||||
namespace mongo {
|
||||
|
||||
class ServiceContext;
|
||||
|
||||
/**
|
||||
* Owns the OpenTelemetry instruments for disk I/O metrics. The set of devices is fixed at
|
||||
* construction time: disks that appear after startup are not tracked.
|
||||
*/
|
||||
class DiskMetrics {
|
||||
public:
|
||||
/**
|
||||
* Registers per-device counters for each of the provided disk devices.
|
||||
*/
|
||||
explicit DiskMetrics(std::vector<std::string> disks);
|
||||
~DiskMetrics();
|
||||
|
||||
/**
|
||||
* Walks the BSON and adds deltas to the registered counters. Devices not declared at
|
||||
* construction time are ignored.
|
||||
*/
|
||||
void update(BSONObj disksBson);
|
||||
|
||||
private:
|
||||
class Impl;
|
||||
std::unique_ptr<Impl> _impl;
|
||||
};
|
||||
|
||||
/**
|
||||
* Registers OpenTelemetry disk I/O counters and starts a periodic job that samples
|
||||
* once per second. No-op on unsupported platforms.
|
||||
*/
|
||||
MONGO_MOD_PUBLIC void installDiskOtelMetrics(ServiceContext* svcCtx);
|
||||
|
||||
} // namespace mongo
|
||||
209
src/mongo/otel/metrics/instrumentation/disk_metrics_linux.cpp
Normal file
209
src/mongo/otel/metrics/instrumentation/disk_metrics_linux.cpp
Normal file
@ -0,0 +1,209 @@
|
||||
/**
|
||||
* Copyright (C) 2026-present MongoDB, Inc.
|
||||
*
|
||||
* This program is free software: you can redistribute it and/or modify
|
||||
* it under the terms of the Server Side Public License, version 1,
|
||||
* as published by MongoDB, Inc.
|
||||
*
|
||||
* This program is distributed in the hope that it will be useful,
|
||||
* but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
||||
* Server Side Public License for more details.
|
||||
*
|
||||
* You should have received a copy of the Server Side Public License
|
||||
* along with this program. If not, see
|
||||
* <http://www.mongodb.com/licensing/server-side-public-license>.
|
||||
*
|
||||
* As a special exception, the copyright holders give permission to link the
|
||||
* code of portions of this program with the OpenSSL library under certain
|
||||
* conditions as described in each individual source file and distribute
|
||||
* linked combinations including the program with the OpenSSL library. You
|
||||
* must comply with the Server Side Public License in all respects for
|
||||
* all of the code used other than as permitted herein. If you modify file(s)
|
||||
* with this exception, you may extend this exception to your version of the
|
||||
* file(s), but you are not obligated to do so. If you do not wish to do so,
|
||||
* delete this exception statement from your version. If you delete this
|
||||
* exception statement from all source files in the program, then also delete
|
||||
* it in the license file.
|
||||
*/
|
||||
|
||||
#include "mongo/base/string_data.h"
|
||||
#include "mongo/bson/bsonobj.h"
|
||||
#include "mongo/bson/bsonobjbuilder.h"
|
||||
#include "mongo/db/service_context.h"
|
||||
#include "mongo/logv2/log.h"
|
||||
#include "mongo/logv2/log_severity_suppressor.h"
|
||||
#include "mongo/otel/metrics/instrumentation/disk_metrics.h"
|
||||
#include "mongo/otel/metrics/metric_unit.h"
|
||||
#include "mongo/otel/metrics/metrics_counter.h"
|
||||
#include "mongo/otel/metrics/metrics_service.h"
|
||||
#include "mongo/util/duration.h"
|
||||
#include "mongo/util/periodic_runner.h"
|
||||
#include "mongo/util/procparser.h"
|
||||
|
||||
#include <utility>
|
||||
#include <vector>
|
||||
|
||||
#include <boost/optional.hpp>
|
||||
#include <fmt/format.h>
|
||||
|
||||
#define MONGO_LOGV2_DEFAULT_COMPONENT ::mongo::logv2::LogComponent::kControl
|
||||
|
||||
namespace mongo {
|
||||
|
||||
namespace {
|
||||
|
||||
using otel::metrics::Counter;
|
||||
using otel::metrics::DynamicMetricNameMaker;
|
||||
using otel::metrics::MetricsService;
|
||||
using otel::metrics::MetricUnit;
|
||||
|
||||
constexpr StringData kDiskStatsPath = "/proc/diskstats"_sd;
|
||||
constexpr StringData kSysBlockPath = "/sys/block"_sd;
|
||||
|
||||
struct DiskCounters {
|
||||
Counter<int64_t>* reads{nullptr};
|
||||
Counter<int64_t>* readSectors{nullptr};
|
||||
Counter<int64_t>* readTimeMs{nullptr};
|
||||
Counter<int64_t>* writes{nullptr};
|
||||
Counter<int64_t>* writeSectors{nullptr};
|
||||
Counter<int64_t>* writeTimeMs{nullptr};
|
||||
Counter<int64_t>* ioTimeMs{nullptr};
|
||||
Counter<int64_t>* ioQueuedMs{nullptr};
|
||||
};
|
||||
|
||||
struct DiskMetricsState {
|
||||
std::unique_ptr<DiskMetrics> metrics;
|
||||
std::vector<std::string> diskNames;
|
||||
std::vector<StringData> diskViews;
|
||||
PeriodicJobAnchor job;
|
||||
};
|
||||
|
||||
const auto getDiskMetricsState = ServiceContext::declareDecoration<DiskMetricsState>();
|
||||
|
||||
} // namespace
|
||||
|
||||
class DiskMetrics::Impl {
|
||||
public:
|
||||
explicit Impl(std::vector<std::string> disks) : _disks(std::move(disks)) {
|
||||
_instruments.resize(_disks.size());
|
||||
|
||||
for (size_t i = 0; i < _disks.size(); ++i) {
|
||||
const std::string& disk = _disks[i];
|
||||
const auto makeCounter =
|
||||
[&](StringData field, std::string desc, MetricUnit unit) -> Counter<int64_t>* {
|
||||
std::string fullName = fmt::format("systemMetrics.disks.{}.{}", disk, field);
|
||||
return &MetricsService::instance().createInt64Counter(
|
||||
DynamicMetricNameMaker::make(StringData{fullName}), std::move(desc), unit);
|
||||
};
|
||||
|
||||
_instruments[i].reads = makeCounter(
|
||||
"reads", "Number of read operations completed", MetricUnit::kOperations);
|
||||
_instruments[i].readSectors =
|
||||
makeCounter("read_sectors", "Number of sectors read", MetricUnit::kCount);
|
||||
_instruments[i].readTimeMs =
|
||||
makeCounter("read_time_ms", "Time spent reading", MetricUnit::kMilliseconds);
|
||||
_instruments[i].writes = makeCounter(
|
||||
"writes", "Number of write operations completed", MetricUnit::kOperations);
|
||||
_instruments[i].writeSectors =
|
||||
makeCounter("write_sectors", "Number of sectors written", MetricUnit::kCount);
|
||||
_instruments[i].writeTimeMs =
|
||||
makeCounter("write_time_ms", "Time spent writing", MetricUnit::kMilliseconds);
|
||||
_instruments[i].ioTimeMs = makeCounter(
|
||||
"io_time_ms", "Time disk was busy doing I/O", MetricUnit::kMilliseconds);
|
||||
_instruments[i].ioQueuedMs = makeCounter("io_queued_ms",
|
||||
"Weighted time spent in the disk I/O queue",
|
||||
MetricUnit::kMilliseconds);
|
||||
}
|
||||
}
|
||||
|
||||
void update(BSONObj disksBson) {
|
||||
if (!_previousBson) {
|
||||
_previousBson = std::move(disksBson);
|
||||
return;
|
||||
}
|
||||
|
||||
for (size_t i = 0; i < _disks.size(); ++i) {
|
||||
const BSONElement prev = (*_previousBson)[_disks[i]];
|
||||
const BSONElement curr = disksBson[_disks[i]];
|
||||
|
||||
if (prev.type() != BSONType::object || curr.type() != BSONType::object) {
|
||||
continue;
|
||||
}
|
||||
|
||||
addDeltas(_instruments[i], prev.Obj(), curr.Obj());
|
||||
}
|
||||
|
||||
_previousBson = std::move(disksBson);
|
||||
}
|
||||
|
||||
private:
|
||||
void addDeltas(DiskCounters& instr, const BSONObj& prev, const BSONObj& curr) {
|
||||
const auto delta = [&](StringData field) {
|
||||
return std::max(0LL, curr[field].safeNumberLong() - prev[field].safeNumberLong());
|
||||
};
|
||||
|
||||
instr.reads->add(delta("reads"));
|
||||
instr.readSectors->add(delta("read_sectors"));
|
||||
instr.readTimeMs->add(delta("read_time_ms"));
|
||||
instr.writes->add(delta("writes"));
|
||||
instr.writeSectors->add(delta("write_sectors"));
|
||||
instr.writeTimeMs->add(delta("write_time_ms"));
|
||||
instr.ioTimeMs->add(delta("io_time_ms"));
|
||||
instr.ioQueuedMs->add(delta("io_queued_ms"));
|
||||
}
|
||||
|
||||
std::vector<std::string> _disks;
|
||||
std::vector<DiskCounters> _instruments;
|
||||
boost::optional<BSONObj> _previousBson;
|
||||
};
|
||||
|
||||
DiskMetrics::DiskMetrics(std::vector<std::string> disks)
|
||||
: _impl(std::make_unique<Impl>(std::move(disks))) {}
|
||||
|
||||
DiskMetrics::~DiskMetrics() = default;
|
||||
|
||||
void DiskMetrics::update(BSONObj disksBson) {
|
||||
_impl->update(std::move(disksBson));
|
||||
}
|
||||
|
||||
void installDiskOtelMetrics(ServiceContext* svcCtx) {
|
||||
auto& state = getDiskMetricsState(svcCtx);
|
||||
|
||||
state.diskNames = procparser::findPhysicalDisks(kSysBlockPath);
|
||||
if (state.diskNames.empty()) {
|
||||
return;
|
||||
}
|
||||
|
||||
state.diskViews.reserve(state.diskNames.size());
|
||||
for (const auto& name : state.diskNames) {
|
||||
state.diskViews.push_back(name);
|
||||
}
|
||||
|
||||
state.metrics = std::make_unique<DiskMetrics>(state.diskNames);
|
||||
|
||||
state.job = svcCtx->getPeriodicRunner()->makeJob(PeriodicRunner::PeriodicJob{
|
||||
"DiskOtelMetrics",
|
||||
[&state](Client*) {
|
||||
BSONObjBuilder builder;
|
||||
Status s =
|
||||
procparser::parseProcDiskStatsFile(kDiskStatsPath, state.diskViews, &builder);
|
||||
if (s.isOK()) {
|
||||
state.metrics->update(builder.obj());
|
||||
} else {
|
||||
static logv2::SeveritySuppressor suppressor(
|
||||
Minutes{1}, logv2::LogSeverity::Warning(), logv2::LogSeverity::Debug(3));
|
||||
|
||||
if (auto sev = suppressor(); shouldLog(MONGO_LOGV2_DEFAULT_COMPONENT, sev)) {
|
||||
LOGV2_DEBUG(
|
||||
12006910, sev.toInt(), "Failed to collect disk stats", "error"_attr = s);
|
||||
}
|
||||
}
|
||||
},
|
||||
Seconds(1),
|
||||
false /*isKillableByStepdown*/});
|
||||
|
||||
state.job.start();
|
||||
}
|
||||
|
||||
} // namespace mongo
|
||||
44
src/mongo/otel/metrics/instrumentation/disk_metrics_noop.cpp
Normal file
44
src/mongo/otel/metrics/instrumentation/disk_metrics_noop.cpp
Normal file
@ -0,0 +1,44 @@
|
||||
/**
|
||||
* Copyright (C) 2026-present MongoDB, Inc.
|
||||
*
|
||||
* This program is free software: you can redistribute it and/or modify
|
||||
* it under the terms of the Server Side Public License, version 1,
|
||||
* as published by MongoDB, Inc.
|
||||
*
|
||||
* This program is distributed in the hope that it will be useful,
|
||||
* but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
||||
* Server Side Public License for more details.
|
||||
*
|
||||
* You should have received a copy of the Server Side Public License
|
||||
* along with this program. If not, see
|
||||
* <http://www.mongodb.com/licensing/server-side-public-license>.
|
||||
*
|
||||
* As a special exception, the copyright holders give permission to link the
|
||||
* code of portions of this program with the OpenSSL library under certain
|
||||
* conditions as described in each individual source file and distribute
|
||||
* linked combinations including the program with the OpenSSL library. You
|
||||
* must comply with the Server Side Public License in all respects for
|
||||
* all of the code used other than as permitted herein. If you modify file(s)
|
||||
* with this exception, you may extend this exception to your version of the
|
||||
* file(s), but you are not obligated to do so. If you do not wish to do so,
|
||||
* delete this exception statement from your version. If you delete this
|
||||
* exception statement from all source files in the program, then also delete
|
||||
* it in the license file.
|
||||
*/
|
||||
|
||||
#include "mongo/otel/metrics/instrumentation/disk_metrics.h"
|
||||
|
||||
namespace mongo {
|
||||
|
||||
class DiskMetrics::Impl {};
|
||||
|
||||
DiskMetrics::DiskMetrics(std::vector<std::string>) : _impl(std::make_unique<Impl>()) {}
|
||||
|
||||
DiskMetrics::~DiskMetrics() = default;
|
||||
|
||||
void DiskMetrics::update(BSONObj) {}
|
||||
|
||||
void installDiskOtelMetrics(ServiceContext*) {}
|
||||
|
||||
} // namespace mongo
|
||||
250
src/mongo/otel/metrics/instrumentation/disk_metrics_test.cpp
Normal file
250
src/mongo/otel/metrics/instrumentation/disk_metrics_test.cpp
Normal file
@ -0,0 +1,250 @@
|
||||
/**
|
||||
* Copyright (C) 2026-present MongoDB, Inc.
|
||||
*
|
||||
* This program is free software: you can redistribute it and/or modify
|
||||
* it under the terms of the Server Side Public License, version 1,
|
||||
* as published by MongoDB, Inc.
|
||||
*
|
||||
* This program is distributed in the hope that it will be useful,
|
||||
* but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
||||
* Server Side Public License for more details.
|
||||
*
|
||||
* You should have received a copy of the Server Side Public License
|
||||
* along with this program. If not, see
|
||||
* <http://www.mongodb.com/licensing/server-side-public-license>.
|
||||
*
|
||||
* As a special exception, the copyright holders give permission to link the
|
||||
* code of portions of this program with the OpenSSL library under certain
|
||||
* conditions as described in each individual source file and distribute
|
||||
* linked combinations including the program with the OpenSSL library. You
|
||||
* must comply with the Server Side Public License in all respects for
|
||||
* all of the code used other than as permitted herein. If you modify file(s)
|
||||
* with this exception, you may extend this exception to your version of the
|
||||
* file(s), but you are not obligated to do so. If you do not wish to do so,
|
||||
* delete this exception statement from your version. If you delete this
|
||||
* exception statement from all source files in the program, then also delete
|
||||
* it in the license file.
|
||||
*/
|
||||
|
||||
#include "mongo/otel/metrics/instrumentation/disk_metrics.h"
|
||||
|
||||
#include "mongo/bson/bsonobjbuilder.h"
|
||||
#include "mongo/otel/metrics/metrics_test_util.h"
|
||||
#include "mongo/unittest/unittest.h"
|
||||
|
||||
namespace mongo {
|
||||
namespace {
|
||||
|
||||
using otel::metrics::DynamicMetricNameMaker;
|
||||
using otel::metrics::OtelMetricsCapturer;
|
||||
|
||||
constexpr StringData kSda = "sda"_sd;
|
||||
constexpr StringData kSdb = "sdb"_sd;
|
||||
constexpr StringData kSdc = "sdc"_sd;
|
||||
|
||||
// All tests register DiskMetrics with this fixed device set. The global MetricsService retains
|
||||
// registrations across tests within the same binary, so every test must use the same device set.
|
||||
const std::vector<std::string> kTestDevices = {std::string(kSda), std::string(kSdb)};
|
||||
|
||||
constexpr StringData kSdaReads = "systemMetrics.disks.sda.reads"_sd;
|
||||
constexpr StringData kSdaReadSectors = "systemMetrics.disks.sda.read_sectors"_sd;
|
||||
constexpr StringData kSdaReadTimeMs = "systemMetrics.disks.sda.read_time_ms"_sd;
|
||||
constexpr StringData kSdaWrites = "systemMetrics.disks.sda.writes"_sd;
|
||||
constexpr StringData kSdaWriteSectors = "systemMetrics.disks.sda.write_sectors"_sd;
|
||||
constexpr StringData kSdaWriteTimeMs = "systemMetrics.disks.sda.write_time_ms"_sd;
|
||||
constexpr StringData kSdaIoTimeMs = "systemMetrics.disks.sda.io_time_ms"_sd;
|
||||
constexpr StringData kSdaIoQueuedMs = "systemMetrics.disks.sda.io_queued_ms"_sd;
|
||||
constexpr StringData kSdbWrites = "systemMetrics.disks.sdb.writes"_sd;
|
||||
constexpr StringData kSdcWrites = "systemMetrics.disks.sdc.writes"_sd;
|
||||
|
||||
BSONObj makeDiskBson(StringData device,
|
||||
long long reads,
|
||||
long long readSectors,
|
||||
long long readTimeMs,
|
||||
long long writes,
|
||||
long long writeSectors,
|
||||
long long writeTimeMs,
|
||||
long long ioTimeMs,
|
||||
long long ioQueuedMs) {
|
||||
BSONObjBuilder b;
|
||||
{
|
||||
BSONObjBuilder sub(b.subobjStart(device));
|
||||
sub.appendNumber("reads", reads);
|
||||
sub.appendNumber("read_sectors", readSectors);
|
||||
sub.appendNumber("read_time_ms", readTimeMs);
|
||||
sub.appendNumber("writes", writes);
|
||||
sub.appendNumber("write_sectors", writeSectors);
|
||||
sub.appendNumber("write_time_ms", writeTimeMs);
|
||||
sub.appendNumber("io_time_ms", ioTimeMs);
|
||||
sub.appendNumber("io_queued_ms", ioQueuedMs);
|
||||
}
|
||||
return b.obj();
|
||||
}
|
||||
|
||||
class DiskOtelMetricsTest : public unittest::Test {
|
||||
protected:
|
||||
void setUp() override {
|
||||
if (!OtelMetricsCapturer::canReadMetrics()) {
|
||||
GTEST_SKIP() << "Skipping test: OTel metrics unavailable on this platform";
|
||||
}
|
||||
}
|
||||
|
||||
OtelMetricsCapturer _capturer;
|
||||
DiskMetrics _metrics{kTestDevices};
|
||||
};
|
||||
|
||||
TEST_F(DiskOtelMetricsTest, FirstUpdateSetsBaseline) {
|
||||
ASSERT_DOES_NOT_THROW(
|
||||
_metrics.update(makeDiskBson(kSda, 10, 20, 50, 50, 100, 300, 1000, 1200)));
|
||||
}
|
||||
|
||||
TEST_F(DiskOtelMetricsTest, SecondUpdateEmitsDeltas) {
|
||||
_metrics.update(makeDiskBson(kSda,
|
||||
/*reads=*/10,
|
||||
/*readSectors=*/20,
|
||||
/*readTimeMs=*/50,
|
||||
/*writes=*/50,
|
||||
/*writeSectors=*/100,
|
||||
/*writeTimeMs=*/300,
|
||||
/*ioTimeMs=*/1000,
|
||||
/*ioQueuedMs=*/1200));
|
||||
|
||||
_metrics.update(makeDiskBson(kSda,
|
||||
/*reads=*/25,
|
||||
/*readSectors=*/50,
|
||||
/*readTimeMs=*/120,
|
||||
/*writes=*/70,
|
||||
/*writeSectors=*/130,
|
||||
/*writeTimeMs=*/500,
|
||||
/*ioTimeMs=*/1500,
|
||||
/*ioQueuedMs=*/1800));
|
||||
|
||||
ASSERT_EQ(_capturer.readInt64Counter(DynamicMetricNameMaker::make(kSdaReads)), 15);
|
||||
ASSERT_EQ(_capturer.readInt64Counter(DynamicMetricNameMaker::make(kSdaReadSectors)), 30);
|
||||
ASSERT_EQ(_capturer.readInt64Counter(DynamicMetricNameMaker::make(kSdaReadTimeMs)), 70);
|
||||
ASSERT_EQ(_capturer.readInt64Counter(DynamicMetricNameMaker::make(kSdaWrites)), 20);
|
||||
ASSERT_EQ(_capturer.readInt64Counter(DynamicMetricNameMaker::make(kSdaWriteSectors)), 30);
|
||||
ASSERT_EQ(_capturer.readInt64Counter(DynamicMetricNameMaker::make(kSdaWriteTimeMs)), 200);
|
||||
ASSERT_EQ(_capturer.readInt64Counter(DynamicMetricNameMaker::make(kSdaIoTimeMs)), 500);
|
||||
ASSERT_EQ(_capturer.readInt64Counter(DynamicMetricNameMaker::make(kSdaIoQueuedMs)), 600);
|
||||
}
|
||||
|
||||
TEST_F(DiskOtelMetricsTest, MultipleDeltasAccumulate) {
|
||||
_metrics.update(makeDiskBson(kSda, 50, 100, 200, 100, 200, 500, 1000, 1500));
|
||||
_metrics.update(makeDiskBson(kSda, 60, 120, 260, 105, 210, 560, 1100, 1650));
|
||||
_metrics.update(makeDiskBson(kSda, 80, 160, 380, 115, 230, 680, 1300, 1950));
|
||||
|
||||
// reads delta: (60-50) + (80-60) = 10 + 20 = 30
|
||||
ASSERT_EQ(_capturer.readInt64Counter(DynamicMetricNameMaker::make(kSdaReads)), 30);
|
||||
// read_sectors delta: (120-100) + (160-120) = 20 + 40 = 60
|
||||
ASSERT_EQ(_capturer.readInt64Counter(DynamicMetricNameMaker::make(kSdaReadSectors)), 60);
|
||||
// read_time_ms delta: (260-200) + (380-260) = 60 + 120 = 180
|
||||
ASSERT_EQ(_capturer.readInt64Counter(DynamicMetricNameMaker::make(kSdaReadTimeMs)), 180);
|
||||
// writes delta: (105-100) + (115-105) = 5 + 10 = 15
|
||||
ASSERT_EQ(_capturer.readInt64Counter(DynamicMetricNameMaker::make(kSdaWrites)), 15);
|
||||
// write_sectors delta: (210-200) + (230-210) = 10 + 20 = 30
|
||||
ASSERT_EQ(_capturer.readInt64Counter(DynamicMetricNameMaker::make(kSdaWriteSectors)), 30);
|
||||
// write_time_ms delta: (560-500) + (680-560) = 60 + 120 = 180
|
||||
ASSERT_EQ(_capturer.readInt64Counter(DynamicMetricNameMaker::make(kSdaWriteTimeMs)), 180);
|
||||
// io_time_ms delta: (1100-1000) + (1300-1100) = 100 + 200 = 300
|
||||
ASSERT_EQ(_capturer.readInt64Counter(DynamicMetricNameMaker::make(kSdaIoTimeMs)), 300);
|
||||
// io_queued_ms delta: (1650-1500) + (1950-1650) = 150 + 300 = 450
|
||||
ASSERT_EQ(_capturer.readInt64Counter(DynamicMetricNameMaker::make(kSdaIoQueuedMs)), 450);
|
||||
}
|
||||
|
||||
TEST_F(DiskOtelMetricsTest, UndeclaredDevicesAreIgnored) {
|
||||
_metrics.update(makeDiskBson(kSda, 0, 0, 0, 0, 0, 0, 0, 0));
|
||||
|
||||
BSONObjBuilder both;
|
||||
{
|
||||
BSONObjBuilder sub(both.subobjStart(kSda));
|
||||
sub.appendNumber("reads", 10LL);
|
||||
sub.appendNumber("read_sectors", 20LL);
|
||||
sub.appendNumber("read_time_ms", 50LL);
|
||||
sub.appendNumber("writes", 20LL);
|
||||
sub.appendNumber("write_sectors", 40LL);
|
||||
sub.appendNumber("write_time_ms", 100LL);
|
||||
sub.appendNumber("io_time_ms", 200LL);
|
||||
sub.appendNumber("io_queued_ms", 300LL);
|
||||
}
|
||||
{
|
||||
// sdc was never registered — its data must be ignored.
|
||||
BSONObjBuilder sub(both.subobjStart(kSdc));
|
||||
sub.appendNumber("reads", 100LL);
|
||||
sub.appendNumber("read_sectors", 200LL);
|
||||
sub.appendNumber("read_time_ms", 500LL);
|
||||
sub.appendNumber("writes", 500LL);
|
||||
sub.appendNumber("write_sectors", 1000LL);
|
||||
sub.appendNumber("write_time_ms", 3000LL);
|
||||
sub.appendNumber("io_time_ms", 5000LL);
|
||||
sub.appendNumber("io_queued_ms", 7000LL);
|
||||
}
|
||||
_metrics.update(both.obj());
|
||||
|
||||
ASSERT_EQ(_capturer.readInt64Counter(DynamicMetricNameMaker::make(kSdaWrites)), 20);
|
||||
|
||||
// sdc was never registered, so its metric name does not exist in the service.
|
||||
ASSERT_THROWS_CODE(_capturer.readInt64Counter(DynamicMetricNameMaker::make(kSdcWrites)),
|
||||
DBException,
|
||||
ErrorCodes::KeyNotFound);
|
||||
}
|
||||
|
||||
TEST_F(DiskOtelMetricsTest, MultipleRegisteredDevicesTrackedIndependently) {
|
||||
BSONObjBuilder baseline;
|
||||
{
|
||||
BSONObjBuilder sub(baseline.subobjStart(kSda));
|
||||
sub.appendNumber("reads", 40LL);
|
||||
sub.appendNumber("read_sectors", 80LL);
|
||||
sub.appendNumber("read_time_ms", 200LL);
|
||||
sub.appendNumber("writes", 100LL);
|
||||
sub.appendNumber("write_sectors", 200LL);
|
||||
sub.appendNumber("write_time_ms", 500LL);
|
||||
sub.appendNumber("io_time_ms", 1000LL);
|
||||
sub.appendNumber("io_queued_ms", 1500LL);
|
||||
}
|
||||
{
|
||||
BSONObjBuilder sub(baseline.subobjStart(kSdb));
|
||||
sub.appendNumber("reads", 20LL);
|
||||
sub.appendNumber("read_sectors", 40LL);
|
||||
sub.appendNumber("read_time_ms", 100LL);
|
||||
sub.appendNumber("writes", 50LL);
|
||||
sub.appendNumber("write_sectors", 80LL);
|
||||
sub.appendNumber("write_time_ms", 200LL);
|
||||
sub.appendNumber("io_time_ms", 400LL);
|
||||
sub.appendNumber("io_queued_ms", 600LL);
|
||||
}
|
||||
_metrics.update(baseline.obj());
|
||||
|
||||
BSONObjBuilder next;
|
||||
{
|
||||
BSONObjBuilder sub(next.subobjStart(kSda));
|
||||
sub.appendNumber("reads", 50LL);
|
||||
sub.appendNumber("read_sectors", 100LL);
|
||||
sub.appendNumber("read_time_ms", 250LL);
|
||||
sub.appendNumber("writes", 110LL);
|
||||
sub.appendNumber("write_sectors", 220LL);
|
||||
sub.appendNumber("write_time_ms", 560LL);
|
||||
sub.appendNumber("io_time_ms", 1100LL);
|
||||
sub.appendNumber("io_queued_ms", 1650LL);
|
||||
}
|
||||
{
|
||||
BSONObjBuilder sub(next.subobjStart(kSdb));
|
||||
sub.appendNumber("reads", 23LL);
|
||||
sub.appendNumber("read_sectors", 46LL);
|
||||
sub.appendNumber("read_time_ms", 115LL);
|
||||
sub.appendNumber("writes", 55LL);
|
||||
sub.appendNumber("write_sectors", 90LL);
|
||||
sub.appendNumber("write_time_ms", 230LL);
|
||||
sub.appendNumber("io_time_ms", 440LL);
|
||||
sub.appendNumber("io_queued_ms", 660LL);
|
||||
}
|
||||
_metrics.update(next.obj());
|
||||
|
||||
ASSERT_EQ(_capturer.readInt64Counter(DynamicMetricNameMaker::make(kSdaReads)), 10);
|
||||
ASSERT_EQ(_capturer.readInt64Counter(DynamicMetricNameMaker::make(kSdaWrites)), 10);
|
||||
ASSERT_EQ(_capturer.readInt64Counter(DynamicMetricNameMaker::make(kSdbWrites)), 5);
|
||||
}
|
||||
|
||||
} // namespace
|
||||
} // namespace mongo
|
||||
@ -0,0 +1,72 @@
|
||||
/**
|
||||
* Copyright (C) 2026-present MongoDB, Inc.
|
||||
*
|
||||
* This program is free software: you can redistribute it and/or modify
|
||||
* it under the terms of the Server Side Public License, version 1,
|
||||
* as published by MongoDB, Inc.
|
||||
*
|
||||
* This program is distributed in the hope that it will be useful,
|
||||
* but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
||||
* Server Side Public License for more details.
|
||||
*
|
||||
* You should have received a copy of the Server Side Public License
|
||||
* along with this program. If not, see
|
||||
* <http://www.mongodb.com/licensing/server-side-public-license>.
|
||||
*
|
||||
* As a special exception, the copyright holders give permission to link the
|
||||
* code of portions of this program with the OpenSSL library under certain
|
||||
* conditions as described in each individual source file and distribute
|
||||
* linked combinations including the program with the OpenSSL library. You
|
||||
* must comply with the Server Side Public License in all respects for
|
||||
* all of the code used other than as permitted herein. If you modify file(s)
|
||||
* with this exception, you may extend this exception to your version of the
|
||||
* file(s), but you are not obligated to do so. If you do not wish to do so,
|
||||
* delete this exception statement from your version. If you delete this
|
||||
* exception statement from all source files in the program, then also delete
|
||||
* it in the license file.
|
||||
*/
|
||||
|
||||
#pragma once
|
||||
|
||||
#include "mongo/bson/bsonobj.h"
|
||||
#include "mongo/util/modules.h"
|
||||
|
||||
#include <memory>
|
||||
#include <string>
|
||||
#include <vector>
|
||||
|
||||
namespace mongo {
|
||||
|
||||
class ServiceContext;
|
||||
|
||||
/**
|
||||
* Owns the OpenTelemetry instruments for filesystem mount metrics. The set of mountpoints is
|
||||
* fixed at construction time and mounts that appear after startup are not tracked.
|
||||
*/
|
||||
class SystemMountMetrics {
|
||||
public:
|
||||
/**
|
||||
* Registers per-mountpoint gauges for each of the provided mountpoints.
|
||||
*/
|
||||
explicit SystemMountMetrics(std::vector<std::string> mountpoints);
|
||||
~SystemMountMetrics();
|
||||
|
||||
/**
|
||||
* Walks the BSON and pushes the values to the registered gauges. Mountpoints not declared at
|
||||
* construction time are ignored.
|
||||
*/
|
||||
void update(const BSONObj& mountsBson);
|
||||
|
||||
private:
|
||||
class Impl;
|
||||
std::unique_ptr<Impl> _impl;
|
||||
};
|
||||
|
||||
/**
|
||||
* Registers OpenTelemetry mount filesystem gauges and starts a periodic job that samples
|
||||
* OS-level mount stats once per second. No-op on unsupported platforms.
|
||||
*/
|
||||
MONGO_MOD_PUBLIC void installSystemMountOtelMetrics(ServiceContext* svcCtx);
|
||||
|
||||
} // namespace mongo
|
||||
@ -0,0 +1,218 @@
|
||||
/**
|
||||
* Copyright (C) 2026-present MongoDB, Inc.
|
||||
*
|
||||
* This program is free software: you can redistribute it and/or modify
|
||||
* it under the terms of the Server Side Public License, version 1,
|
||||
* as published by MongoDB, Inc.
|
||||
*
|
||||
* This program is distributed in the hope that it will be useful,
|
||||
* but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
||||
* Server Side Public License for more details.
|
||||
*
|
||||
* You should have received a copy of the Server Side Public License
|
||||
* along with this program. If not, see
|
||||
* <http://www.mongodb.com/licensing/server-side-public-license>.
|
||||
*
|
||||
* As a special exception, the copyright holders give permission to link the
|
||||
* code of portions of this program with the OpenSSL library under certain
|
||||
* conditions as described in each individual source file and distribute
|
||||
* linked combinations including the program with the OpenSSL library. You
|
||||
* must comply with the Server Side Public License in all respects for
|
||||
* all of the code used other than as permitted herein. If you modify file(s)
|
||||
* with this exception, you may extend this exception to your version of the
|
||||
* file(s), but you are not obligated to do so. If you do not wish to do so,
|
||||
* delete this exception statement from your version. If you delete this
|
||||
* exception statement from all source files in the program, then also delete
|
||||
* it in the license file.
|
||||
*/
|
||||
|
||||
#include "mongo/base/string_data.h"
|
||||
#include "mongo/bson/bsonobjbuilder.h"
|
||||
#include "mongo/db/service_context.h"
|
||||
#include "mongo/logv2/log.h"
|
||||
#include "mongo/logv2/log_severity_suppressor.h"
|
||||
#include "mongo/otel/metrics/instrumentation/system_mount_metrics.h"
|
||||
#include "mongo/otel/metrics/metric_unit.h"
|
||||
#include "mongo/otel/metrics/metrics_gauge.h"
|
||||
#include "mongo/otel/metrics/metrics_service.h"
|
||||
#include "mongo/util/assert_util.h"
|
||||
#include "mongo/util/duration.h"
|
||||
#include "mongo/util/periodic_runner.h"
|
||||
#include "mongo/util/procparser.h"
|
||||
|
||||
#include <algorithm>
|
||||
#include <iterator>
|
||||
#include <string>
|
||||
#include <utility>
|
||||
#include <vector>
|
||||
|
||||
#include <fmt/format.h>
|
||||
|
||||
#define MONGO_LOGV2_DEFAULT_COMPONENT ::mongo::logv2::LogComponent::kControl
|
||||
|
||||
namespace mongo {
|
||||
|
||||
namespace {
|
||||
|
||||
using otel::metrics::DynamicMetricNameMaker;
|
||||
using otel::metrics::Gauge;
|
||||
using otel::metrics::MetricsService;
|
||||
using otel::metrics::MetricUnit;
|
||||
|
||||
struct MountGauges {
|
||||
Gauge<int64_t>* capacity{nullptr};
|
||||
Gauge<int64_t>* available{nullptr};
|
||||
Gauge<int64_t>* free{nullptr};
|
||||
};
|
||||
|
||||
struct MountOtelMetricsState {
|
||||
std::unique_ptr<SystemMountMetrics> metrics;
|
||||
PeriodicJobAnchor job;
|
||||
};
|
||||
|
||||
const auto getMountOtelMetricsState = ServiceContext::declareDecoration<MountOtelMetricsState>();
|
||||
|
||||
constexpr StringData kMountInfoPath = "/proc/self/mountinfo"_sd;
|
||||
|
||||
// Sanitize a mount path for use as a metric name segment:
|
||||
// "/" -> "root"
|
||||
// "/data" -> "data"
|
||||
// "/boot/efi" -> "boot.efi"
|
||||
std::string sanitizeMountpoint(StringData path) {
|
||||
if (path == "/") {
|
||||
return "root";
|
||||
}
|
||||
|
||||
// Strip leading slash and replace interior slashes with dots.
|
||||
std::string result(path.substr(1));
|
||||
std::replace(result.begin(), result.end(), '/', '.');
|
||||
return result;
|
||||
}
|
||||
|
||||
std::vector<std::string> enumerateMountpoints() {
|
||||
BSONObjBuilder probe;
|
||||
|
||||
const Status status = procparser::parseProcSelfMountStatsFile(kMountInfoPath, &probe);
|
||||
|
||||
if (!status.isOK()) {
|
||||
LOGV2_WARNING(12006900,
|
||||
"Failed to enumerate mountpoints for OTel system mount metrics",
|
||||
"error"_attr = status);
|
||||
return {};
|
||||
}
|
||||
|
||||
const BSONObj probeObj = probe.obj();
|
||||
std::vector<std::string> mountpoints;
|
||||
std::ranges::transform(probeObj, std::back_inserter(mountpoints), [](const BSONElement& entry) {
|
||||
return std::string(entry.fieldName());
|
||||
});
|
||||
|
||||
std::ranges::sort(mountpoints);
|
||||
mountpoints.erase(std::ranges::unique(mountpoints).begin(), mountpoints.end());
|
||||
|
||||
return mountpoints;
|
||||
}
|
||||
|
||||
std::unique_ptr<SystemMountMetrics> makeMetrics() {
|
||||
auto mountpoints = enumerateMountpoints();
|
||||
if (mountpoints.empty()) {
|
||||
return nullptr;
|
||||
}
|
||||
|
||||
try {
|
||||
return std::make_unique<SystemMountMetrics>(std::move(mountpoints));
|
||||
} catch (const DBException& ex) {
|
||||
LOGV2_ERROR(
|
||||
12006901, "Failed to register OTel system mount metrics", "error"_attr = ex.toStatus());
|
||||
return nullptr;
|
||||
}
|
||||
}
|
||||
|
||||
} // namespace
|
||||
|
||||
class SystemMountMetrics::Impl {
|
||||
public:
|
||||
explicit Impl(std::vector<std::string> mountpoints) : _mountpoints(std::move(mountpoints)) {
|
||||
_instruments.resize(_mountpoints.size());
|
||||
|
||||
for (size_t i = 0; i < _mountpoints.size(); ++i) {
|
||||
const auto& mountpoint = _mountpoints[i];
|
||||
const auto sanitized = sanitizeMountpoint(mountpoint);
|
||||
|
||||
const auto makeGauge =
|
||||
[&](StringData field, std::string desc, MetricUnit unit) -> Gauge<int64_t>* {
|
||||
std::string fullName = fmt::format("systemMetrics.mounts.{}.{}", sanitized, field);
|
||||
return &MetricsService::instance().createInt64Gauge(
|
||||
DynamicMetricNameMaker::make(StringData{fullName}), std::move(desc), unit);
|
||||
};
|
||||
|
||||
_instruments[i].capacity =
|
||||
makeGauge("capacity", "Total filesystem capacity in bytes", MetricUnit::kBytes);
|
||||
_instruments[i].available =
|
||||
makeGauge("available", "Filesystem space available in bytes", MetricUnit::kBytes);
|
||||
_instruments[i].free =
|
||||
makeGauge("free", "Total free filesystem space in bytes", MetricUnit::kBytes);
|
||||
}
|
||||
}
|
||||
|
||||
void update(const BSONObj& mountsBson) {
|
||||
for (size_t i = 0; i < _mountpoints.size(); ++i) {
|
||||
const BSONElement entry = mountsBson[_mountpoints[i]];
|
||||
if (entry.type() != BSONType::object) {
|
||||
continue;
|
||||
}
|
||||
|
||||
const BSONObj stats = entry.Obj();
|
||||
_instruments[i].capacity->set(stats["capacity"].safeNumberLong());
|
||||
_instruments[i].available->set(stats["available"].safeNumberLong());
|
||||
_instruments[i].free->set(stats["free"].safeNumberLong());
|
||||
}
|
||||
}
|
||||
|
||||
private:
|
||||
std::vector<std::string> _mountpoints;
|
||||
std::vector<MountGauges> _instruments;
|
||||
};
|
||||
|
||||
SystemMountMetrics::SystemMountMetrics(std::vector<std::string> mountpoints)
|
||||
: _impl(std::make_unique<Impl>(std::move(mountpoints))) {}
|
||||
|
||||
SystemMountMetrics::~SystemMountMetrics() = default;
|
||||
|
||||
void SystemMountMetrics::update(const BSONObj& mountsBson) {
|
||||
_impl->update(mountsBson);
|
||||
}
|
||||
|
||||
void installSystemMountOtelMetrics(ServiceContext* svcCtx) {
|
||||
auto metrics = makeMetrics();
|
||||
if (!metrics) {
|
||||
return;
|
||||
}
|
||||
|
||||
auto& state = getMountOtelMetricsState(svcCtx);
|
||||
state.metrics = std::move(metrics);
|
||||
state.job = svcCtx->getPeriodicRunner()->makeJob(PeriodicRunner::PeriodicJob{
|
||||
"SystemMountOtelMetrics",
|
||||
[&state](Client*) {
|
||||
BSONObjBuilder builder;
|
||||
Status s = procparser::parseProcSelfMountStatsFile(kMountInfoPath, &builder);
|
||||
if (s.isOK()) {
|
||||
state.metrics->update(builder.obj());
|
||||
} else {
|
||||
static logv2::SeveritySuppressor suppressor(
|
||||
Minutes{1}, logv2::LogSeverity::Warning(), logv2::LogSeverity::Debug(3));
|
||||
|
||||
if (auto sev = suppressor(); shouldLog(MONGO_LOGV2_DEFAULT_COMPONENT, sev)) {
|
||||
LOGV2_DEBUG(
|
||||
12006902, sev.toInt(), "Failed to collect mount stats", "error"_attr = s);
|
||||
}
|
||||
}
|
||||
},
|
||||
Seconds(1),
|
||||
false /*isKillableByStepdown*/});
|
||||
|
||||
state.job.start();
|
||||
}
|
||||
|
||||
} // namespace mongo
|
||||
@ -0,0 +1,45 @@
|
||||
/**
|
||||
* Copyright (C) 2026-present MongoDB, Inc.
|
||||
*
|
||||
* This program is free software: you can redistribute it and/or modify
|
||||
* it under the terms of the Server Side Public License, version 1,
|
||||
* as published by MongoDB, Inc.
|
||||
*
|
||||
* This program is distributed in the hope that it will be useful,
|
||||
* but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
||||
* Server Side Public License for more details.
|
||||
*
|
||||
* You should have received a copy of the Server Side Public License
|
||||
* along with this program. If not, see
|
||||
* <http://www.mongodb.com/licensing/server-side-public-license>.
|
||||
*
|
||||
* As a special exception, the copyright holders give permission to link the
|
||||
* code of portions of this program with the OpenSSL library under certain
|
||||
* conditions as described in each individual source file and distribute
|
||||
* linked combinations including the program with the OpenSSL library. You
|
||||
* must comply with the Server Side Public License in all respects for
|
||||
* all of the code used other than as permitted herein. If you modify file(s)
|
||||
* with this exception, you may extend this exception to your version of the
|
||||
* file(s), but you are not obligated to do so. If you do not wish to do so,
|
||||
* delete this exception statement from your version. If you delete this
|
||||
* exception statement from all source files in the program, then also delete
|
||||
* it in the license file.
|
||||
*/
|
||||
|
||||
#include "mongo/otel/metrics/instrumentation/system_mount_metrics.h"
|
||||
|
||||
namespace mongo {
|
||||
|
||||
class SystemMountMetrics::Impl {};
|
||||
|
||||
SystemMountMetrics::SystemMountMetrics(std::vector<std::string>)
|
||||
: _impl(std::make_unique<Impl>()) {}
|
||||
|
||||
SystemMountMetrics::~SystemMountMetrics() = default;
|
||||
|
||||
void SystemMountMetrics::update(const BSONObj&) {}
|
||||
|
||||
void installSystemMountOtelMetrics(ServiceContext*) {}
|
||||
|
||||
} // namespace mongo
|
||||
@ -0,0 +1,124 @@
|
||||
/**
|
||||
* Copyright (C) 2026-present MongoDB, Inc.
|
||||
*
|
||||
* This program is free software: you can redistribute it and/or modify
|
||||
* it under the terms of the Server Side Public License, version 1,
|
||||
* as published by MongoDB, Inc.
|
||||
*
|
||||
* This program is distributed in the hope that it will be useful,
|
||||
* but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
||||
* Server Side Public License for more details.
|
||||
*
|
||||
* You should have received a copy of the Server Side Public License
|
||||
* along with this program. If not, see
|
||||
* <http://www.mongodb.com/licensing/server-side-public-license>.
|
||||
*
|
||||
* As a special exception, the copyright holders give permission to link the
|
||||
* code of portions of this program with the OpenSSL library under certain
|
||||
* conditions as described in each individual source file and distribute
|
||||
* linked combinations including the program with the OpenSSL library. You
|
||||
* must comply with the Server Side Public License in all respects for
|
||||
* all of the code used other than as permitted herein. If you modify file(s)
|
||||
* with this exception, you may extend this exception to your version of the
|
||||
* file(s), but you are not obligated to do so. If you do not wish to do so,
|
||||
* delete this exception statement from your version. If you delete this
|
||||
* exception statement from all source files in the program, then also delete
|
||||
* it in the license file.
|
||||
*/
|
||||
|
||||
#include "mongo/otel/metrics/instrumentation/system_mount_metrics.h"
|
||||
|
||||
#include "mongo/bson/bsonobjbuilder.h"
|
||||
#include "mongo/otel/metrics/metrics_test_util.h"
|
||||
#include "mongo/unittest/unittest.h"
|
||||
|
||||
namespace mongo {
|
||||
namespace {
|
||||
|
||||
using otel::metrics::DynamicMetricNameMaker;
|
||||
using otel::metrics::OtelMetricsCapturer;
|
||||
|
||||
constexpr StringData kDataMount = "/data"_sd;
|
||||
constexpr StringData kTmpMount = "/tmp"_sd;
|
||||
|
||||
// All tests register SystemMountMetrics with this fixed mountpoint set. The global MetricsService
|
||||
// retains registrations across tests within the same binary, so every test must use the same set.
|
||||
const std::vector<std::string> kTestMountpoints = {std::string(kDataMount)};
|
||||
|
||||
constexpr StringData kDataCapacity = "systemMetrics.mounts.data.capacity"_sd;
|
||||
constexpr StringData kDataAvailable = "systemMetrics.mounts.data.available"_sd;
|
||||
constexpr StringData kDataFree = "systemMetrics.mounts.data.free"_sd;
|
||||
constexpr StringData kTmpCapacity = "systemMetrics.mounts.tmp.capacity"_sd;
|
||||
|
||||
BSONObj makeMountsBson(StringData mountpoint,
|
||||
long long capacity,
|
||||
long long available,
|
||||
long long free) {
|
||||
BSONObjBuilder b;
|
||||
{
|
||||
BSONObjBuilder sub(b.subobjStart(mountpoint));
|
||||
sub.appendNumber("capacity", capacity);
|
||||
sub.appendNumber("available", available);
|
||||
sub.appendNumber("free", free);
|
||||
}
|
||||
return b.obj();
|
||||
}
|
||||
|
||||
class SystemMountOtelMetricsTest : public unittest::Test {
|
||||
protected:
|
||||
void setUp() override {
|
||||
if (!OtelMetricsCapturer::canReadMetrics()) {
|
||||
GTEST_SKIP() << "Skipping test: OTel metrics unavailable on this platform";
|
||||
}
|
||||
}
|
||||
|
||||
OtelMetricsCapturer _capturer;
|
||||
SystemMountMetrics _metrics{kTestMountpoints};
|
||||
};
|
||||
|
||||
TEST_F(SystemMountOtelMetricsTest, UpdateSetsGaugeValues) {
|
||||
_metrics.update(makeMountsBson(kDataMount, 1000, 400, 500));
|
||||
|
||||
ASSERT_EQ(_capturer.readInt64Gauge(DynamicMetricNameMaker::make(kDataCapacity)), 1000);
|
||||
ASSERT_EQ(_capturer.readInt64Gauge(DynamicMetricNameMaker::make(kDataAvailable)), 400);
|
||||
ASSERT_EQ(_capturer.readInt64Gauge(DynamicMetricNameMaker::make(kDataFree)), 500);
|
||||
}
|
||||
|
||||
TEST_F(SystemMountOtelMetricsTest, UpdateIgnoresUndeclaredMountpoints) {
|
||||
_metrics.update(makeMountsBson(kDataMount, 1000, 400, 500));
|
||||
|
||||
BSONObjBuilder both;
|
||||
{
|
||||
BSONObjBuilder sub(both.subobjStart(kDataMount));
|
||||
sub.appendNumber("capacity", 2000LL);
|
||||
sub.appendNumber("available", 800LL);
|
||||
sub.appendNumber("free", 1000LL);
|
||||
}
|
||||
{
|
||||
// /tmp was never registered — its data must be ignored.
|
||||
BSONObjBuilder sub(both.subobjStart(kTmpMount));
|
||||
sub.appendNumber("capacity", 500LL);
|
||||
sub.appendNumber("available", 200LL);
|
||||
sub.appendNumber("free", 250LL);
|
||||
}
|
||||
_metrics.update(both.obj());
|
||||
|
||||
ASSERT_EQ(_capturer.readInt64Gauge(DynamicMetricNameMaker::make(kDataCapacity)), 2000);
|
||||
|
||||
// /tmp was never registered, so its metric name does not exist in the service.
|
||||
ASSERT_THROWS_CODE(_capturer.readInt64Gauge(DynamicMetricNameMaker::make(kTmpCapacity)),
|
||||
DBException,
|
||||
ErrorCodes::KeyNotFound);
|
||||
}
|
||||
|
||||
TEST_F(SystemMountOtelMetricsTest, UpdateIsIdempotentForSameValues) {
|
||||
BSONObj bson = makeMountsBson(kDataMount, 8000, 3000, 4000);
|
||||
_metrics.update(bson);
|
||||
_metrics.update(bson);
|
||||
|
||||
ASSERT_EQ(_capturer.readInt64Gauge(DynamicMetricNameMaker::make(kDataCapacity)), 8000);
|
||||
}
|
||||
|
||||
} // namespace
|
||||
} // namespace mongo
|
||||
@ -60,7 +60,6 @@ public:
|
||||
*/
|
||||
constexpr MetricName(StringData name, Passkey<MetricNameMaker>) : _name(name) {}
|
||||
constexpr MetricName(StringData name, Passkey<disagg::MetricNameMaker>) : _name(name) {}
|
||||
|
||||
constexpr StringData getName() const {
|
||||
return _name;
|
||||
}
|
||||
@ -79,6 +78,19 @@ class MONGO_MOD_FILE_PRIVATE MetricNameMaker{public : static constexpr MetricNam
|
||||
} // namespace otel::metrics
|
||||
}; // namespace mongo
|
||||
|
||||
/**
|
||||
* Helper to create MetricName instances with runtime-constructed names (e.g. names that embed
|
||||
* device names or mount paths discovered at startup). Requires N&O review since dynamic names
|
||||
* cannot be audited at compile time.
|
||||
*
|
||||
* TODO(SERVER-127521): Ensure ServerStatusOptions is boost::none for any runtime metric.
|
||||
*/
|
||||
class MONGO_MOD_PUBLIC DynamicMetricNameMaker{
|
||||
public : static MetricName make(StringData name){return MetricNameMaker::make(name);
|
||||
}
|
||||
}
|
||||
;
|
||||
|
||||
/**
|
||||
* Central registry of OpenTelemetry metric names used in the server. When adding a new metric to
|
||||
* the server, please add an entry to MetricNames grouped under your team name.
|
||||
|
||||
Loading…
Reference in New Issue
Block a user