SERVER-61914: add fault facet details to FaultImpl::toBSON

SERVER-61958: Make a mocked health observer class that takes in failpoint data

13 files changed, 244 insertions, 21 deletions

diff --git a/jstests/sharding/health_monitor/server_status_health.js b/jstests/sharding/health_monitor/server_status_health.js
index a0d24c704fd..76244deddb7 100644
--- a/jstests/sharding/health_monitor/server_status_health.js
+++ b/jstests/sharding/health_monitor/server_status_health.js
@@ -1,18 +1,75 @@
-
-
+/**
+ * Tests server status has correct fault/facet information.
+ */
 (function() {
 'use strict';
 
-var st = new ShardingTest({
-    mongos: 1,
+const params = {
+    setParameter: {
+        healthMonitoring: tojson({test: "off", ldap: "off", dns: "off"}),
+        featureFlagHealthMonitoring: true
+    }
+};
+
+let st = new ShardingTest({
+    mongos: [params],
     shards: 1,
     other: {mongosOptions: {setParameter: {featureFlagHealthMonitoring: true}}}
 });
 
-const result = assert.commandWorked(st.s0.adminCommand({serverStatus: 1})).health;
+// Check server status after initial health check is complete.
+let result = assert.commandWorked(st.s0.adminCommand({serverStatus: 1})).health;
+print("---RESULT 1---");
 print(tojson(result));
+assert.eq(result.state, "Ok");
+assert(result.enteredStateAtTime);
+
+assert.commandWorked(st.s0.adminCommand(
+    {"setParameter": 1, healthMonitoring: {test: "critical", dns: 'off', ldap: 'off'}}));
 
-assert(result.state == "StartupCheck" || result.state == "Ok");
+// Check server status after test health observer enabled and failpoint returns fault.
+assert.commandWorked(st.s0.adminCommand({
+    "configureFailPoint": 'testHealthObserver',
+    "data": {"code": "InternalError", "msg": "test msg"},
+    "mode": "alwaysOn"
+}));
+
+assert.soon(() => {
+    result = assert.commandWorked(st.s0.adminCommand({serverStatus: 1})).health;
+    return result.state == "TransientFault";
+});
+
+print("---RESULT 2---");
+print(tojson(result));
+assert(result.enteredStateAtTime);
+assert(result.faultInformation);
+
+const faultInformation = result.faultInformation;
+assert.eq(faultInformation.severity, 1);
+assert(faultInformation.duration);
+assert(faultInformation.facets);
+assert.eq(faultInformation.numFacets, 1);
+assert(faultInformation.facets.kTestObserver);
+
+const kTestObserverFacet = faultInformation.facets.kTestObserver;
+assert.eq(kTestObserverFacet.severity, faultInformation.severity);
+assert.eq(kTestObserverFacet.duration, faultInformation.duration);
+assert(kTestObserverFacet.description.includes("InternalError: test msg"));
+
+// Check server status after test health observer enabled and failpoint returns success.
+assert.commandWorked(
+    st.s0.adminCommand({"configureFailPoint": 'testHealthObserver', "mode": "alwaysOn"}));
+
+assert.soon(() => {
+    result = assert.commandWorked(st.s0.adminCommand({serverStatus: 1})).health;
+    return result.state == "Ok";
+});
+
+result = assert.commandWorked(st.s0.adminCommand({serverStatus: 1})).health;
+print("---RESULT 3---");
+print(tojson(result));
+assert.eq(result.state, "Ok");
+assert(result.enteredStateAtTime);
 
 st.stop();
 })();
diff --git a/src/mongo/db/process_health/SConscript b/src/mongo/db/process_health/SConscript
index 82bdded2855..ce02fc06253 100644
--- a/src/mongo/db/process_health/SConscript
+++ b/src/mongo/db/process_health/SConscript
@@ -17,6 +17,7 @@ env.Library(
         'health_observer_base.cpp',
         'health_observer_registration.cpp',
         'progress_monitor.cpp',
+        'test_health_observer.cpp',
     ],
     LIBDEPS=[
         '$BUILD_DIR/mongo/base',
diff --git a/src/mongo/db/process_health/fault_facet.h b/src/mongo/db/process_health/fault_facet.h
index bab42ef4d5e..1927921b254 100644
--- a/src/mongo/db/process_health/fault_facet.h
+++ b/src/mongo/db/process_health/fault_facet.h
@@ -53,10 +53,20 @@ public:
      */
     virtual HealthCheckStatus getStatus() const = 0;
 
+    virtual Milliseconds getDuration() const = 0;
+
     /**
      * Change the state of this Facet with health check result.
      */
     virtual void update(HealthCheckStatus status) = 0;
+
+    virtual void appendDescription(BSONObjBuilder* builder) const = 0;
+
+    BSONObj toBSON() const {
+        BSONObjBuilder builder;
+        appendDescription(&builder);
+        return builder.obj();
+    }
 };
 
 using FaultFacetPtr = std::shared_ptr<FaultFacet>;
diff --git a/src/mongo/db/process_health/fault_facet_impl.cpp b/src/mongo/db/process_health/fault_facet_impl.cpp
index f2975b6b559..62d6c4585ba 100644
--- a/src/mongo/db/process_health/fault_facet_impl.cpp
+++ b/src/mongo/db/process_health/fault_facet_impl.cpp
@@ -48,11 +48,22 @@ HealthCheckStatus FaultFacetImpl::getStatus() const {
     return HealthCheckStatus(getType(), _severity, _description);
 }
 
+Milliseconds FaultFacetImpl::getDuration() const {
+    return std::max(Milliseconds(0), Milliseconds(_clockSource->now() - _startTime));
+}
+
 void FaultFacetImpl::update(HealthCheckStatus status) {
     auto lk = stdx::lock_guard(_mutex);
     _severity = status.getSeverity();
     _description = status.getShortDescription().toString();
 }
 
+void FaultFacetImpl::appendDescription(BSONObjBuilder* builder) const {
+    builder->append("type", FaultFacetType_serializer(getType()));
+    builder->append("severity", _severity);
+    builder->append("duration", getDuration().toBSON());
+    builder->append("description", _description);
+};
+
 }  // namespace process_health
 }  // namespace mongo
diff --git a/src/mongo/db/process_health/fault_facet_impl.h b/src/mongo/db/process_health/fault_facet_impl.h
index 76279052470..aa48b2343f5 100644
--- a/src/mongo/db/process_health/fault_facet_impl.h
+++ b/src/mongo/db/process_health/fault_facet_impl.h
@@ -50,8 +50,12 @@ public:
 
     HealthCheckStatus getStatus() const override;
 
+    Milliseconds getDuration() const override;
+
     void update(HealthCheckStatus status) override;
 
+    void appendDescription(BSONObjBuilder* builder) const override;
+
 private:
     const FaultFacetType _type;
     ClockSource* const _clockSource;
diff --git a/src/mongo/db/process_health/fault_facet_mock.h b/src/mongo/db/process_health/fault_facet_mock.h
index 6035a5c89df..29a3c3a4184 100644
--- a/src/mongo/db/process_health/fault_facet_mock.h
+++ b/src/mongo/db/process_health/fault_facet_mock.h
@@ -67,6 +67,15 @@ public:
         return healthCheckStatus;
     }
 
+    Milliseconds getDuration() const override {
+        return std::max(Milliseconds(0), Milliseconds(_clockSource->now() - _startTime));
+    }
+
+    void appendDescription(BSONObjBuilder* builder) const override {
+        builder->append("type", FaultFacetType_serializer(getType()));
+        builder->append("duration", getDuration().toBSON());
+    };
+
     void update(HealthCheckStatus status) override {
         MONGO_UNREACHABLE;  // Don't use this in mock.
     }
diff --git a/src/mongo/db/process_health/fault_impl.cpp b/src/mongo/db/process_health/fault_impl.cpp
index 77ca9a82cc5..0fe95df330b 100644
--- a/src/mongo/db/process_health/fault_impl.cpp
+++ b/src/mongo/db/process_health/fault_impl.cpp
@@ -120,8 +120,13 @@ void FaultImpl::appendDescription(BSONObjBuilder* builder) const {
     builder->append("id", getId().toBSON());
     builder->append("severity", getSeverity());
     builder->append("duration", getDuration().toBSON());
-    // TODO (SERVER-61914): Add fault facet details
-    builder->append("facets", static_cast<int>(_facets.size()));
+    BSONObjBuilder facetsBuilder;
+    for (auto& facet : _facets) {
+        facetsBuilder.append(FaultFacetType_serializer(facet->getType()), facet->toBSON());
+    }
+
+    builder->append("facets", facetsBuilder.obj());
+    builder->append("numFacets", static_cast<int>(_facets.size()));
 }
 
 }  // namespace process_health
diff --git a/src/mongo/db/process_health/fault_manager.cpp b/src/mongo/db/process_health/fault_manager.cpp
index 2dadbfdd2b1..10c43033689 100644
--- a/src/mongo/db/process_health/fault_manager.cpp
+++ b/src/mongo/db/process_health/fault_manager.cpp
@@ -327,6 +327,11 @@ void FaultManager::schedulePeriodicHealthCheckThread() {
         return;
     }
 
+    if (getActiveHealthObservers().size() == 0) {
+        LOGV2_DEBUG(5936511, 2, "No active health observers are configured.");
+        setState(FaultState::kOk, HealthCheckStatus(FaultFacetType::kSystem));
+    }
+
     auto observers = getHealthObservers();
     for (auto observer : observers) {
         LOGV2_DEBUG(
@@ -334,9 +339,7 @@ void FaultManager::schedulePeriodicHealthCheckThread() {
 
         // TODO (SERVER-59368): The system should properly handle a health checker being turned
         // on/off
-        if (_config->isHealthObserverEnabled(observer->getType())) {
-            healthCheck(observer, _managerShuttingDownCancellation);
-        }
+        healthCheck(observer, _managerShuttingDownCancellation);
     }
 }
 
@@ -379,13 +382,7 @@ SharedSemiFuture<void> FaultManager::startPeriodicHealthChecks() {
     invariant(state() == FaultState::kStartupCheck);
 
     _init();
-
-    if (getActiveHealthObservers().size() == 0) {
-        LOGV2_DEBUG(5936511, 2, "No active health observers are configured.");
-        setState(FaultState::kOk, HealthCheckStatus(FaultFacetType::kSystem));
-    } else {
-        schedulePeriodicHealthCheckThread();
-    }
+    schedulePeriodicHealthCheckThread();
 
     return _initialHealthCheckCompletedPromise.getFuture();
 }
@@ -472,6 +469,8 @@ void FaultManager::healthCheck(HealthObserver* observer, std::shared_ptr<AtomicW
         return healthCheckStatus;
     };
 
+    _healthCheckContexts.insert({observer->getType(), HealthCheckContext(nullptr, boost::none)});
+
     // If health observer is disabled, then do nothing and schedule another run (health observer may
     // become enabled).
     // TODO (SERVER-59368): The system should properly handle a health checker being turned on/off
@@ -480,7 +479,6 @@ void FaultManager::healthCheck(HealthObserver* observer, std::shared_ptr<AtomicW
         return;
     }
 
-    _healthCheckContexts.insert({observer->getType(), HealthCheckContext(nullptr, boost::none)});
     // Run asynchronous health check.  When complete, check for state transition (and perform if
     // necessary). Then schedule the next run.
     auto healthCheckFuture = observer->periodicCheck(*this, _taskExecutor, token)
diff --git a/src/mongo/db/process_health/fault_manager_config.h b/src/mongo/db/process_health/fault_manager_config.h
index ce17c74b7a8..92834acafad 100644
--- a/src/mongo/db/process_health/fault_manager_config.h
+++ b/src/mongo/db/process_health/fault_manager_config.h
@@ -63,8 +63,13 @@ std::ostream& operator<<(std::ostream& os, const FaultState& state);
 /**
  * Types of health observers available.
  */
-enum class FaultFacetType { kSystem, kMock1, kMock2, kLdap, kDns };
+enum class FaultFacetType { kSystem, kMock1, kMock2, kTestObserver, kLdap, kDns };
+static const StringData FaultFacetTypeStrings[] = {
+    "kSystem", "kMock1", "kMock2", "kTestObserver", "kLdap", "kDns"};
 
+static const StringData FaultFacetType_serializer(const FaultFacetType value) {
+    return FaultFacetTypeStrings[static_cast<int>(value)];
+}
 
 class FaultManagerConfig {
 public:
@@ -84,6 +89,8 @@ public:
             case FaultFacetType::kDns:
                 return intensities->_data->getDns();
                 // TODO: update this function with additional fault facets when they are added
+            case FaultFacetType::kTestObserver:
+                return intensities->_data->getTest();
             case FaultFacetType::kSystem:
                 return HealthObserverIntensityEnum::kCritical;
             case FaultFacetType::kMock1:
diff --git a/src/mongo/db/process_health/health_check_status.h b/src/mongo/db/process_health/health_check_status.h
index da5284db13c..957432284eb 100644
--- a/src/mongo/db/process_health/health_check_status.h
+++ b/src/mongo/db/process_health/health_check_status.h
@@ -126,7 +126,7 @@ inline StringBuilder& operator<<(StringBuilder& s, const FaultFacetType& type) {
         case FaultFacetType::kMock2:
             return s << "kMock2"_sd;
         default:
-            return s << "Uknown"_sd;
+            return s << "Unknown"_sd;
     }
 }
 
diff --git a/src/mongo/db/process_health/health_monitoring_server_parameters.idl b/src/mongo/db/process_health/health_monitoring_server_parameters.idl
index 1222242a812..42adb48c2d3 100644
--- a/src/mongo/db/process_health/health_monitoring_server_parameters.idl
+++ b/src/mongo/db/process_health/health_monitoring_server_parameters.idl
@@ -55,6 +55,10 @@ structs:
                 description: "Intensity of LDAP fault facet"
                 type: HealthObserverIntensity
                 default: kNonCritical
+            test:
+                description: "Intensity of test fault facet"
+                type: HealthObserverIntensity
+                default: kOff
 
 server_parameters:
     healthMonitoring:
diff --git a/src/mongo/db/process_health/test_health_observer.cpp b/src/mongo/db/process_health/test_health_observer.cpp
new file mode 100644
index 00000000000..2251f9590d0
--- /dev/null
+++ b/src/mongo/db/process_health/test_health_observer.cpp
@@ -0,0 +1,65 @@
+/**
+ *    Copyright (C) 2021-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/init.h"
+#include "mongo/db/process_health/test_health_observer.h"
+#include "mongo/db/process_health/health_observer_registration.h"
+#include "mongo/util/fail_point.h"
+
+namespace mongo {
+namespace process_health {
+MONGO_FAIL_POINT_DEFINE(hangTestHealthObserver);
+MONGO_FAIL_POINT_DEFINE(testHealthObserver);
+Future<HealthCheckStatus> TestHealthObserver::periodicCheckImpl(
+    PeriodicHealthCheckContext&& periodicCheckContext) {
+    hangTestHealthObserver.pauseWhileSet();
+
+    auto result = Future<HealthCheckStatus>::makeReady(makeHealthyStatus());
+
+    testHealthObserver.executeIf(
+        [this, &result](const BSONObj& data) {
+            auto code = data["code"].checkAndGetStringData();
+            auto msg = data["msg"].checkAndGetStringData();
+            result = Future<HealthCheckStatus>::makeReady(makeSimpleFailedStatus(
+                1.0, {Status(ErrorCodes::fromString(code.toString()), msg.toString())}));
+        },
+        [&](const BSONObj& data) { return !data.isEmpty(); });
+
+    return result;
+}
+
+namespace {
+MONGO_INITIALIZER(TestHealthObserver)(InitializerContext*) {
+    HealthObserverRegistration::registerObserverFactory(
+        [](ServiceContext* svcCtx) { return std::make_unique<TestHealthObserver>(svcCtx); });
+    return Status::OK();
+}
+}  // namespace
+}  // namespace process_health
+}  // namespace mongo
diff --git a/src/mongo/db/process_health/test_health_observer.h b/src/mongo/db/process_health/test_health_observer.h
new file mode 100644
index 00000000000..42670ee79a9
--- /dev/null
+++ b/src/mongo/db/process_health/test_health_observer.h
@@ -0,0 +1,52 @@
+/**
+ *    Copyright (C) 2021-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/db/process_health/health_observer_base.h"
+
+namespace mongo {
+namespace process_health {
+class TestHealthObserver : public HealthObserverBase {
+public:
+    FaultFacetType getType() const override {
+        return FaultFacetType::kTestObserver;
+    }
+
+    Milliseconds healthCheckJitter() const override {
+        return Milliseconds(0);
+    }
+
+    TestHealthObserver(ServiceContext* svcCtx) : HealthObserverBase(svcCtx){};
+
+protected:
+    Future<HealthCheckStatus> periodicCheckImpl(
+        PeriodicHealthCheckContext&& periodicCheckContext) override;
+};
+}  // namespace process_health
+}  // namespace mongo