subprocess tracking using Jobs on Windows/hub (#796)HEADmain

This change introduces job object support on Windows to be able to more accurately track and limit resource usage on storage instances created by the hub service. It also ensures that all child instances can be torn down reliably on exit.

Also made it so hub tests no longer pop up console windows while running.

1 files changed, 46 insertions, 3 deletions

diff --git a/src/zenserver/hub/hubservice.cpp b/src/zenserver/hub/hubservice.cpp
index a00446a75..bf0e294c5 100644
--- a/src/zenserver/hub/hubservice.cpp
+++ b/src/zenserver/hub/hubservice.cpp
@@ -8,6 +8,7 @@
 #include <zencore/filesystem.h>
 #include <zencore/fmtutils.h>
 #include <zencore/logging.h>
+#include <zencore/process.h>
 #include <zencore/scopeguard.h>
 #include <zencore/system.h>
 #include <zenutil/zenserverprocess.h>
@@ -150,6 +151,10 @@ struct StorageServerInstance
 
 	inline uint16_t GetBasePort() const { return m_ServerInstance.GetBasePort(); }
 
+#if ZEN_PLATFORM_WINDOWS
+	void SetJobObject(JobObject* InJobObject) { m_JobObject = InJobObject; }
+#endif
+
 private:
 	void				  WakeLocked();
 	RwLock				  m_Lock;
@@ -161,6 +166,9 @@ private:
 	std::filesystem::path m_TempDir;
 	std::filesystem::path m_HydrationPath;
 	ResourceMetrics		  m_ResourceMetrics;
+#if ZEN_PLATFORM_WINDOWS
+	JobObject* m_JobObject = nullptr;
+#endif
 
 	void SpawnServerProcess();
 
@@ -191,6 +199,9 @@ StorageServerInstance::SpawnServerProcess()
 
 	m_ServerInstance.SetServerExecutablePath(GetRunningExecutablePath());
 	m_ServerInstance.SetDataDir(m_BaseDir);
+#if ZEN_PLATFORM_WINDOWS
+	m_ServerInstance.SetJobObject(m_JobObject);
+#endif
 	const uint16_t BasePort = m_ServerInstance.SpawnServerAndWaitUntilReady();
 
 	ZEN_DEBUG("Storage server instance for module '{}' started, listening on port {}", m_ModuleId, BasePort);
@@ -380,6 +391,21 @@ struct HttpHubService::Impl
 		// flexibility, and to allow running multiple hubs on the same host if
 		// necessary.
 		m_RunEnvironment.SetNextPortNumber(21000);
+
+#if ZEN_PLATFORM_WINDOWS
+		if (m_UseJobObject)
+		{
+			m_JobObject.Initialize();
+			if (m_JobObject.IsValid())
+			{
+				ZEN_INFO("Job object initialized for hub service child process management");
+			}
+			else
+			{
+				ZEN_WARN("Failed to initialize job object; child processes will not be auto-terminated on hub crash");
+			}
+		}
+#endif
 	}
 
 	void Cleanup()
@@ -422,6 +448,12 @@ struct HttpHubService::Impl
 				IsNewInstance = true;
 				auto NewInstance =
 					std::make_unique<StorageServerInstance>(m_RunEnvironment, ModuleId, m_FileHydrationPath, m_HydrationTempPath);
+#if ZEN_PLATFORM_WINDOWS
+				if (m_JobObject.IsValid())
+				{
+					NewInstance->SetJobObject(&m_JobObject);
+				}
+#endif
 				Instance = NewInstance.get();
 				m_Instances.emplace(std::string(ModuleId), std::move(NewInstance));
 
@@ -579,10 +611,15 @@ struct HttpHubService::Impl
 	inline int GetInstanceLimit() { return m_InstanceLimit; }
 	inline int GetMaxInstanceCount() { return m_MaxInstanceCount; }
 
+	bool m_UseJobObject = true;
+
 private:
-	ZenServerEnvironment													m_RunEnvironment;
-	std::filesystem::path													m_FileHydrationPath;
-	std::filesystem::path													m_HydrationTempPath;
+	ZenServerEnvironment  m_RunEnvironment;
+	std::filesystem::path m_FileHydrationPath;
+	std::filesystem::path m_HydrationTempPath;
+#if ZEN_PLATFORM_WINDOWS
+	JobObject m_JobObject;
+#endif
 	RwLock																	m_Lock;
 	std::unordered_map<std::string, std::unique_ptr<StorageServerInstance>> m_Instances;
 	std::unordered_set<std::string>											m_DeprovisioningModules;
@@ -817,6 +854,12 @@ HttpHubService::~HttpHubService()
 {
 }
 
+void
+HttpHubService::SetUseJobObject(bool Enable)
+{
+	m_Impl->m_UseJobObject = Enable;
+}
+
 const char*
 HttpHubService::BaseUri() const
 {