Checkpointing work

daggyr/libdaggyr/include/daggyr/Server.hpp

@@ -9,8 +9,8 @@
 #include <daggy/executors/task/DaggyRunnerTaskExecutor.hpp>
 #include <daggy/executors/task/ForkingTaskExecutor.hpp>
 #include <daggy/loggers/dag_run/DAGRunLogger.hpp>
+#include <deque>
 #include <filesystem>
-#include <list>
 
 #define DAGGY_REST_HANDLER(func)                    \
   void func(const Pistache::Rest::Request &request, \
@@ -58,6 +58,8 @@ namespace daggy::daggyr {
 
     executors::task::ForkingTaskExecutor executor_;
 
+    using TaskID = std::pair<DAGRunID, std::string>;
+
     struct TaskRecord
     {
       RunState state;
@@ -68,17 +70,21 @@ namespace daggy::daggyr {
     Capacity maxCapacity_;
     Capacity curCapacity_;
 
-    std::mutex pendingGuard_;
-
     struct PendingJob
     {
-      DAGRunID runID;
-      std::string taskName;
-      std::future<AttemptRecord> fut;
+      daggy::executors::task::TaskFuture fut;
       Capacity resourcesUsed;
       bool resolved;
     };
 
-    std::list<PendingJob> pending_;
+    void monitor();
+    std::atomic<bool> running_;
+    std::thread monitorWorker_;
+
+    std::mutex pendingGuard_;
+    std::unordered_map<TaskID, PendingJob> pending_;
+
+    std::mutex resolvedGuard_;
+    std::deque<std::string> resolved_;
   };
 }  // namespace daggy::daggyr