Fix the user del issue, cgroup tools bug, user creation logic

nodemanager/Version.h

@@ -136,6 +136,13 @@ namespace hpc
                             "Set correct permissions on key files",
                         }
                     },
+                    { "1.2.1.8",
+                        {
+                            "Remove the user even if it is logged on",
+                            "Won't overwrite the keys if private key is not created",
+                            "Work around the bug of removing multi-subsys cgroup in old version of CGroup tools",
+                        }
+                    },
                 };
 
                 return versionHistory;

nodemanager/core/Process.cpp

@@ -165,7 +165,7 @@ void* Process::ForkThread(void* arg)
     }
 
 Final:
-    p->ExecuteCommand("/bin/bash", "CleanupTask.sh", p->taskExecutionId, p->processId);
+    p->ExecuteCommandNoCapture("/bin/bash", "CleanupTask.sh", p->taskExecutionId, p->processId);
     p->ExecuteCommand("rm -rf", p->taskFolder);
     p->ended = true;
     p->OnCompleted();

nodemanager/core/Process.h

@@ -90,6 +90,18 @@ namespace hpc
                     return ret;
                 }
 
+                template <typename ... Args>
+                int ExecuteCommandNoCapture(const std::string& cmd,  const Args& ... args)
+                {
+                    std::string output;
+                    std::string cmdLine = String::Join(" ", cmd, args...);
+                    int ret = System::ExecuteCommandOut(output, cmd, args...);
+
+                    Logger::Debug(this->jobId, this->taskId, this->requeueCount, "'{0}', exitCode {1}, output {2}.", cmdLine, ret, output);
+
+                    return ret;
+                }
+
                 static void* ForkThread(void*);
 
                 std::string GetAffinity();

nodemanager/core/RemoteExecutor.cpp

@@ -61,8 +61,9 @@ json::value RemoteExecutor::StartJobAndTask(StartJobAndTaskArgs&& args, const st
             }
 
             bool privateKeyAdded = 0 == System::AddSshKey(userName, args.PrivateKey, "id_rsa");
-            bool publicKeyAdded = 0 == System::AddSshKey(userName, args.PublicKey, "id_rsa.pub");
+            bool publicKeyAdded = privateKeyAdded && (0 == System::AddSshKey(userName, args.PublicKey, "id_rsa.pub"));
-            bool authKeyAdded = 0 == System::AddAuthorizedKey(userName, args.PublicKey);
+
+            bool authKeyAdded = privateKeyAdded && publicKeyAdded && (0 == System::AddAuthorizedKey(userName, args.PublicKey));
 
             if (authKeyAdded)
             {

nodemanager/utils/System.cpp

@@ -410,7 +410,7 @@ int System::DeleteUser(const std::string& userName)
 {
     std::string output;
 
-    int ret = System::ExecuteCommandOut(output, "userdel", userName, "-r");
+    int ret = System::ExecuteCommandOut(output, "userdel", userName, "-r -f");
     if (ret != 0)
     {
         Logger::Error("userdel {0} error code {1}", userName, ret);