Dev/licenses

.gitignore

@@ -19,7 +19,8 @@ dependencies/.extracted
 .idea
 
 .idea/misc.xml
-
+build/
+.cache/
 # Deployment stuff. Containing some personal info.
 .idea/deployment.xml
 .idea/other.xml

etc/config.yaml

@@ -112,6 +112,9 @@ PendingQueueMaxSize: 900000
 # Default value is 100000.
 ScheduledBatchSize: 100000
 
+# The set licenses and their quantities (Only need to configure in the ctld config file.)
+# Licenses: fluent:30,ansys:100
+
 # Scheduler will reject all jobs beyond processing capacity set by PendingQueueMaxSize
 # if this option is set to true.
 # Default value is false.

protos/Crane.proto

@@ -59,7 +59,7 @@ message SubmitBatchTaskRequest {
 
 message SubmitBatchTaskReply {
   bool ok = 1;
-  oneof payload{
+  oneof payload {
     uint32 task_id = 2;
     string reason = 3;
   }
@@ -90,14 +90,14 @@ message CreateCgroupForTasksRequest {
   repeated string execution_node = 4;
 }
 
-message CreateCgroupForTasksReply{}
+message CreateCgroupForTasksReply {}
 
-message ReleaseCgroupForTasksRequest{
+message ReleaseCgroupForTasksRequest {
   repeated uint32 task_id_list = 1;
   repeated uint32 uid_list = 2;
 }
 
-message ReleaseCgroupForTasksReply{}
+message ReleaseCgroupForTasksReply {}
 
 message TerminateTasksRequest {
   repeated uint32 task_id_list = 1;
@@ -154,30 +154,28 @@ message CancelTaskReply {
   repeated string not_cancelled_reasons = 4;
 }
 
-
-message QueryTaskIdFromPortRequest{
+message QueryTaskIdFromPortRequest {
   uint32 port = 1;
 }
 
-message QueryTaskIdFromPortReply{
+message QueryTaskIdFromPortReply {
   bool ok = 1;
   uint32 task_id = 2;
 }
 
-message QueryTaskIdFromPortForwardRequest{
+message QueryTaskIdFromPortForwardRequest {
   uint32 ssh_remote_port = 1;
   string ssh_remote_address = 2;
   uint32 uid = 3;
 }
 
-message QueryTaskIdFromPortForwardReply{
+message QueryTaskIdFromPortForwardReply {
   bool ok = 1;
   bool from_user = 2;
   uint32 task_id = 3;
   string cgroup_path = 4;
 }
 
-
 message QueryCranedInfoRequest {
   string craned_name = 1;
 }
@@ -194,6 +192,15 @@ message QueryPartitionInfoReply {
   repeated PartitionInfo partition_info = 1;
 }
 
+message QueryLicensesInfoRequest {
+  repeated string license_name_list = 1;
+}
+
+message QueryLicensesInfoReply {
+  bool ok = 1;
+  repeated LicenseInfo license_info_list = 2;
+}
+
 message ModifyTaskRequest {
   enum TargetAttributes {
     TimeLimit = 0;
@@ -230,7 +237,7 @@ message ModifyCranedStateRequest{
   string reason = 4;
 }
 
-message ModifyCranedStateReply{
+message ModifyCranedStateReply {
   repeated string modified_nodes = 2;
   repeated string not_modified_nodes = 3;
   repeated string not_modified_reasons = 4;
@@ -395,21 +402,21 @@ message MigrateSshProcToCgroupReply {
   bool ok = 1;
 }
 
-message QueryTaskEnvVariablesRequest{
+message QueryTaskEnvVariablesRequest {
   uint32 task_id = 1;
 }
 
-message QueryTaskEnvVariablesReply{
+message QueryTaskEnvVariablesReply {
   bool ok = 1;
   map<string/*name*/, string/*value*/> env_map = 2;
 }
 
-message QueryTaskEnvVariablesForwardRequest{
+message QueryTaskEnvVariablesForwardRequest {
   uint32 task_id = 1;
   string execution_node = 2;
 }
 
-message QueryTaskEnvVariablesForwardReply{
+message QueryTaskEnvVariablesForwardReply {
   bool ok = 1;
   map<string/*name*/, string/*value*/> env_map = 2;
 }
@@ -426,7 +433,7 @@ message QueryClusterInfoReply {
   repeated TrimmedPartitionInfo partitions = 2;
 }
 
-message QueryTasksInfoRequest{
+message QueryTasksInfoRequest {
   repeated uint32 filter_task_ids = 1;
   repeated string filter_partitions = 2;
   uint32 num_limit = 3;
@@ -436,14 +443,15 @@ message QueryTasksInfoRequest{
   repeated TaskStatus filter_task_states = 6;
   repeated string filter_users = 7;
   repeated string filter_accounts = 8;
-  TimeInterval filter_submit_time_interval = 9;
-  TimeInterval filter_start_time_interval = 10;
-  TimeInterval filter_end_time_interval = 11;
+  repeated string filter_licenses = 9;
+  TimeInterval filter_submit_time_interval = 10;
+  TimeInterval filter_start_time_interval = 11;
+  TimeInterval filter_end_time_interval = 12;
 
   bool option_include_completed_tasks = 15;
 }
 
-message QueryTasksInfoReply{
+message QueryTasksInfoReply {
   bool ok = 1;
   repeated TaskInfo task_info_list = 2;
 }
@@ -499,7 +507,7 @@ message StreamCallocReply {
     bool ok = 1;
   }
 
-  CforedReplyType type = 1 ;
+  CforedReplyType type = 1;
 
   oneof payload {
     TaskIdReply payload_task_id_reply = 2;
@@ -672,11 +680,11 @@ message StreamCrunReply {
     bool ok = 1;
   }
 
-  message TaskIOForwardReply{
+  message TaskIOForwardReply {
     string msg = 1;
   }
 
-  CforedCrunReplyType type = 1 ;
+  CforedCrunReplyType type = 1;
 
   oneof payload {
     TaskIdReply payload_task_id_reply = 2;
@@ -754,7 +762,7 @@ service CraneCtld {
   rpc CranedRegister(CranedRegisterRequest) returns (CranedRegisterReply);
 
   /* RPCs called from Cfored */
-  rpc CforedStream(stream StreamCforedRequest) returns(stream StreamCtldReply);
+  rpc CforedStream(stream StreamCforedRequest) returns (stream StreamCtldReply);
 
   /* RPCs called from ccancel */
   rpc CancelTask(CancelTaskRequest) returns (CancelTaskReply);
@@ -765,7 +773,10 @@ service CraneCtld {
 
   /* PRCs called from ccontrol */
   rpc QueryCranedInfo(QueryCranedInfoRequest) returns (QueryCranedInfoReply);
-  rpc QueryPartitionInfo(QueryPartitionInfoRequest) returns (QueryPartitionInfoReply);
+  rpc QueryPartitionInfo(QueryPartitionInfoRequest)
+      returns (QueryPartitionInfoReply);
+  rpc QueryLicensesInfo(QueryLicensesInfoRequest)
+      returns (QueryLicensesInfoReply);
   rpc ModifyTask(ModifyTaskRequest) returns (ModifyTaskReply);
   rpc ModifyNode(ModifyCranedStateRequest) returns (ModifyCranedStateReply);
 
@@ -796,34 +807,47 @@ service CraneCtld {
 }
 
 service Craned {
-  /* ----------------------------------- Called from CraneCtld ---------------------------------------------------- */
-  rpc ExecuteTask(ExecuteTasksRequest) returns(ExecuteTasksReply);
+  /* ----------------------------------- Called from CraneCtld
+   * ---------------------------------------------------- */
+  rpc ExecuteTask(ExecuteTasksRequest) returns (ExecuteTasksReply);
 
-  rpc CheckTaskStatus(CheckTaskStatusRequest) returns(CheckTaskStatusReply);
+  rpc CheckTaskStatus(CheckTaskStatusRequest) returns (CheckTaskStatusReply);
 
-  rpc CreateCgroupForTasks(CreateCgroupForTasksRequest) returns(CreateCgroupForTasksReply);
-  rpc ReleaseCgroupForTasks(ReleaseCgroupForTasksRequest) returns(ReleaseCgroupForTasksReply);
+  rpc CreateCgroupForTasks(CreateCgroupForTasksRequest)
+      returns (CreateCgroupForTasksReply);
+  rpc ReleaseCgroupForTasks(ReleaseCgroupForTasksRequest)
+      returns (ReleaseCgroupForTasksReply);
 
-  rpc QueryCranedRemoteMeta(QueryCranedRemoteMetaRequest) returns(QueryCranedRemoteMetaReply);
+  rpc QueryCranedRemoteMeta(QueryCranedRemoteMetaRequest)
+      returns (QueryCranedRemoteMetaReply);
 
   /*
   If the task is an interactive task, the resource uuid is also revoked.
-   If there's no process in this interactive task, just deallocate all the resources.
-   If there are processes in this interactive task, kill all the processes and deallocate resources.
-  If the task is a batch task, just kill it.
+   If there's no process in this interactive task, just deallocate all the
+  resources. If there are processes in this interactive task, kill all the
+  processes and deallocate resources. If the task is a batch task, just kill it.
   */
   rpc TerminateTasks(TerminateTasksRequest) returns (TerminateTasksReply);
-  rpc TerminateOrphanedTask(TerminateOrphanedTaskRequest) returns (TerminateOrphanedTaskReply);
-  rpc ChangeTaskTimeLimit(ChangeTaskTimeLimitRequest) returns (ChangeTaskTimeLimitReply);
-
-  /* ----------------------------------- Called from Craned  ------------------------------------------------------ */
-  rpc QueryTaskIdFromPort(QueryTaskIdFromPortRequest) returns (QueryTaskIdFromPortReply);
-
-  /* ----------------------------------- Called from Pam Module  --------------------------------------------------- */
-  rpc QueryTaskIdFromPortForward(QueryTaskIdFromPortForwardRequest) returns (QueryTaskIdFromPortForwardReply);
-  rpc MigrateSshProcToCgroup(MigrateSshProcToCgroupRequest) returns (MigrateSshProcToCgroupReply);
-  rpc QueryTaskEnvVariables(QueryTaskEnvVariablesRequest) returns (QueryTaskEnvVariablesReply);
-  rpc QueryTaskEnvVariablesForward(QueryTaskEnvVariablesForwardRequest) returns (QueryTaskEnvVariablesForwardReply);
+  rpc TerminateOrphanedTask(TerminateOrphanedTaskRequest)
+      returns (TerminateOrphanedTaskReply);
+  rpc ChangeTaskTimeLimit(ChangeTaskTimeLimitRequest)
+      returns (ChangeTaskTimeLimitReply);
+
+  /* ----------------------------------- Called from Craned
+   * ------------------------------------------------------ */
+  rpc QueryTaskIdFromPort(QueryTaskIdFromPortRequest)
+      returns (QueryTaskIdFromPortReply);
+
+  /* ----------------------------------- Called from Pam Module
+   * --------------------------------------------------- */
+  rpc QueryTaskIdFromPortForward(QueryTaskIdFromPortForwardRequest)
+      returns (QueryTaskIdFromPortForwardReply);
+  rpc MigrateSshProcToCgroup(MigrateSshProcToCgroupRequest)
+      returns (MigrateSshProcToCgroupReply);
+  rpc QueryTaskEnvVariables(QueryTaskEnvVariablesRequest)
+      returns (QueryTaskEnvVariablesReply);
+  rpc QueryTaskEnvVariablesForward(QueryTaskEnvVariablesForwardRequest)
+      returns (QueryTaskEnvVariablesForwardReply);
 }
 
 service CraneForeD {

protos/PublicDefs.proto

@@ -141,11 +141,13 @@ message TaskToCtld {
   string extra_attr = 23;
 
   string cmd_line = 31;
-  string cwd = 32; // Current working directory
+  string cwd = 32;  // Current working directory
   map<string, string> env = 33;
 
   string excludes = 34;
   string nodelist = 35;
+
+  map<string, uint32> licenses_count = 36;
 }
 
 message TaskInEmbeddedDb {
@@ -183,7 +185,7 @@ message TaskToD {
   // If this task is PENDING, start_time is either not set (default constructed)
   // or an estimated start time.
   // If this task is RUNNING, start_time is the actual starting time.
-  google.protobuf.Timestamp start_time = 5; // Currently Only used in CraneCtld
+  google.protobuf.Timestamp start_time = 5;  // Currently Only used in CraneCtld
   google.protobuf.Duration time_limit = 6;
 
   string partition = 8;
@@ -260,6 +262,7 @@ message TaskInfo {
   string username = 15;
   string qos = 16;
   ResourceView res_view = 17;
+  map<string, uint32> licenses_count = 18;
   repeated string req_nodes = 19;
   repeated string exclude_nodes = 20;
 
@@ -276,10 +279,11 @@ message TaskInfo {
     string craned_list = 36;
   }
 
-  // The time of different nodes across the whole cluster might not always be synchronized.
-  // If the time on the front end node is more than several seconds ahead of the CraneCtld node,
-  // a negative elapsed time might occur.
-  // To avoid this, the elapsed time of a task is calculated on the CraneCtld side.
+  // The time of different nodes across the whole cluster might not always be
+  // synchronized. If the time on the front end node is more than several
+  // seconds ahead of the CraneCtld node, a negative elapsed time might occur.
+  // To avoid this, the elapsed time of a task is calculated on the CraneCtld
+  // side.
   google.protobuf.Duration elapsed_time = 37;
   repeated string execution_node = 38;
 }
@@ -434,15 +438,17 @@ message AccountInfo {
   bool blocked = 10;
 }
 
-// Note: UserInfo DIFFERS from the `User` struct in C++ code and database representation
-//       and is ONLY used for communication between CraneCtld and cacctmgr command.
-//       If an user belongs to multiple accounts, There will be multiple `UserInfo`
-//       messages with `account` pointing to each account.
+// Note: UserInfo DIFFERS from the `User` struct in C++ code and database
+// representation
+//       and is ONLY used for communication between CraneCtld and cacctmgr
+//       command. If an user belongs to multiple accounts, There will be
+//       multiple `UserInfo` messages with `account` pointing to each account.
 //       For example, if a user (uid=1) belongs to accounts `1,2,3`,
-//       there will be three `UserInfo` messages: (uid=1, account=1), (uid=1, account=2),
-//       (uid=1, account=3).
-//       The c++ code and database representation use a Map<account name, AttrsInAccount> to contain
-//       in ONE UserInfo message all the information belonging to different accounts.
+//       there will be three `UserInfo` messages: (uid=1, account=1), (uid=1,
+//       account=2), (uid=1, account=3). The c++ code and database
+//       representation use a Map<account name, AttrsInAccount> to contain in
+//       ONE UserInfo message all the information belonging to different
+//       accounts.
 message UserInfo {
   enum AdminLevel {
     None = 0;
@@ -492,3 +498,9 @@ message CranedRemoteMeta {
   google.protobuf.Timestamp craned_start_time = 4;
   google.protobuf.Timestamp system_boot_time = 5;
 }
+message LicenseInfo {
+  string name = 1;
+  uint32 total = 2;
+  uint32 used = 3;
+  uint32 free = 4;
+}

src/CraneCtld/CMakeLists.txt

@@ -14,6 +14,8 @@ add_executable(cranectld
         AccountManager.cpp
         AccountMetaContainer.h
         AccountMetaContainer.cpp
+        LicensesManager.h
+        LicensesManager.cpp
         EmbeddedDbClient.cpp
         EmbeddedDbClient.h
         CraneCtld.cpp