fix(cmapi): MCOL-6091 CMAPI gives DMLProc only 10 seconds for a greceful stop

Fix default timeout for stop node, stop dml proc, shutdown Controller, put_config handler etc.
All places that could cause reducing dmlproc graceful stop timeout are fixed:
- cluster and node shutdown
- stop_dmlproc
- start_transaction
- put_config
- toggle_cluster_state

cmapi/cmapi_server/constants.py

@@ -114,7 +114,10 @@ MCS_BACKUP_MANAGER_SH = os.path.join(MCS_INSTALL_BIN, 'mcs_backup_manager.sh')
 CMAPI_PORT = 8640  #TODO: use it in all places
 CURRENT_NODE_CMAPI_URL = f'https://localhost:{CMAPI_PORT}'
 REQUEST_TIMEOUT: float = 30.0
-TRANSACTION_TIMEOUT: float = 300.0  # 5 minutes
+
+DMLPROC_SHUTDOWN_TIMEOUT: float = 300.0  # 5 minutes, should be less then LONG_REQUEST_TIMEOUT
+LONG_REQUEST_TIMEOUT: float = 400.0  # should be less than TRANSACTION_TIMEOUT
+TRANSACTION_TIMEOUT: float = 600.0  # 10 minutes
 
 # API version
 _version = '0.4.0'

cmapi/cmapi_server/controllers/endpoints.py

@@ -16,10 +16,18 @@ from mcs_node_control.models.node_status import NodeStatus
 from pydantic import ValidationError
 
 from cmapi_server.constants import (
-    CMAPI_PACKAGE_NAME, CMAPI_PORT, DEFAULT_MCS_CONF_PATH,
+    CMAPI_PACKAGE_NAME,
-    DEFAULT_SM_CONF_PATH, EM_PATH_SUFFIX, MCS_BRM_CURRENT_PATH, MCS_EM_PATH,
+    CMAPI_PORT,
-    MDB_CS_PACKAGE_NAME, MDB_SERVER_PACKAGE_NAME, REQUEST_TIMEOUT,
+    DEFAULT_MCS_CONF_PATH,
-    S3_BRM_CURRENT_PATH, SECRET_KEY,
+    DMLPROC_SHUTDOWN_TIMEOUT,
+    EM_PATH_SUFFIX,
+    MCS_BRM_CURRENT_PATH,
+    MCS_EM_PATH,
+    MDB_CS_PACKAGE_NAME,
+    MDB_SERVER_PACKAGE_NAME,
+    REQUEST_TIMEOUT,
+    S3_BRM_CURRENT_PATH,
+    SECRET_KEY,
 )
 from cmapi_server.controllers.api_clients import NodeControllerClient
 from cmapi_server.controllers.error import APIError
@@ -725,7 +733,7 @@ class ShutdownController:
         req = cherrypy.request
         use_sudo = get_use_sudo(req.app.config)
         request_body = cherrypy.request.json
-        timeout = request_body.get('timeout', 0)
+        timeout = request_body.get('timeout', DMLPROC_SHUTDOWN_TIMEOUT)
         node_config = NodeConfig()
         try:
             MCSProcessManager.stop_node(
@@ -894,7 +902,7 @@ class ClusterController:
 
         request = cherrypy.request
         request_body = request.json
-        timeout = request_body.get('timeout', None)
+        timeout = request_body.get('timeout', DMLPROC_SHUTDOWN_TIMEOUT)
         force = request_body.get('force', False)
         config = request_body.get('config', DEFAULT_MCS_CONF_PATH)
         in_transaction = request_body.get('in_transaction', False)
@@ -904,7 +912,7 @@ class ClusterController:
                 with TransactionManager():
                     response = ClusterHandler.shutdown(config, timeout)
             else:
-                response = ClusterHandler.shutdown(config)
+                response = ClusterHandler.shutdown(config, timeout)
         except CMAPIBasicError as err:
             raise_422_error(module_logger, func_name, err.message)
 
@@ -1594,7 +1602,7 @@ class NodeProcessController():
 
         request = cherrypy.request
         request_body = request.json
-        timeout = request_body.get('timeout', 10)
+        timeout = request_body.get('timeout', DMLPROC_SHUTDOWN_TIMEOUT)
         force = request_body.get('force', False)
 
         if force:

cmapi/cmapi_server/handlers/cluster.py

@@ -16,7 +16,11 @@ from mcs_node_control.models.node_config import NodeConfig
 from tracing.traced_session import get_traced_session
 
 from cmapi_server.constants import (
-    CMAPI_CONF_PATH, CMAPI_PORT, DEFAULT_MCS_CONF_PATH, REQUEST_TIMEOUT,
+    CMAPI_CONF_PATH,
+    CMAPI_PORT,
+    DEFAULT_MCS_CONF_PATH,
+    DMLPROC_SHUTDOWN_TIMEOUT,
+    REQUEST_TIMEOUT,
 )
 from cmapi_server.exceptions import CMAPIBasicError, exc_to_cmapi_error
 from cmapi_server.controllers.api_clients import NodeControllerClient
@@ -44,7 +48,7 @@ class ClusterAction(Enum):
 
 
 def toggle_cluster_state(
-        action: ClusterAction, config: str) -> dict:
+        action: ClusterAction, config: str, timeout: int = DMLPROC_SHUTDOWN_TIMEOUT) -> dict:
     """Toggle the state of the cluster (start or stop).
 
     :param action: The cluster action to perform.
@@ -64,7 +68,7 @@ def toggle_cluster_state(
 
     switch_node_maintenance(maintainance_flag)
     update_revision_and_manager()
-    broadcast_new_config(config, distribute_secrets=True)
+    broadcast_new_config(config, distribute_secrets=True, timeout=timeout)
 
 
 class ClusterHandler:
@@ -161,7 +165,7 @@ class ClusterHandler:
 
     @staticmethod
     def shutdown(
-        config: str = DEFAULT_MCS_CONF_PATH, timeout: Optional[int] = None
+        config: str = DEFAULT_MCS_CONF_PATH, timeout: int = DMLPROC_SHUTDOWN_TIMEOUT,
     ) -> dict:
         """Method to stop the MCS Cluster.
 
@@ -169,7 +173,7 @@ class ClusterHandler:
                        defaults to DEFAULT_MCS_CONF_PATH
         :type config: str, optional
         :param timeout: timeout in seconds to gracefully stop DMLProc,
-                        defaults to None
+                        defaults to DMLPROC_SHUTDOWN_TIMEOUT
         :type timeout: Optional[int], optional
         :raises CMAPIBasicError: if no nodes in the cluster
         :return: start timestamp
@@ -180,7 +184,7 @@ class ClusterHandler:
             'Cluster shutdown command called. Shutting down the cluster.'
         )
         operation_start_time = str(datetime.now())
-        toggle_cluster_state(ClusterAction.STOP, config)
+        toggle_cluster_state(ClusterAction.STOP, config, timeout=timeout)
         logger.debug('Successfully finished shutting down the cluster.')
         return {'timestamp': operation_start_time}
 

cmapi/cmapi_server/helpers.py

@@ -27,8 +27,15 @@ from cmapi_server.exceptions import CMAPIBasicError
 requests.packages.urllib3.disable_warnings()  # pylint: disable=no-member
 
 from cmapi_server.constants import (
-    CMAPI_CONF_PATH, CMAPI_DEFAULT_CONF_PATH, DEFAULT_MCS_CONF_PATH,
+    CMAPI_CONF_PATH,
-    DEFAULT_SM_CONF_PATH, LOCALHOSTS, _version
+    CMAPI_DEFAULT_CONF_PATH,
+    DEFAULT_MCS_CONF_PATH,
+    DEFAULT_SM_CONF_PATH,
+    DMLPROC_SHUTDOWN_TIMEOUT,
+    LOCALHOSTS,
+    LONG_REQUEST_TIMEOUT,
+    TRANSACTION_TIMEOUT,
+    _version
 )
 from cmapi_server.handlers.cej import CEJPasswordHandler
 from cmapi_server.managers.process import MCSProcessManager
@@ -54,7 +61,7 @@ def start_transaction(
     remove_nodes: Optional[list] = None,
     optional_nodes: Optional[list] = None,
     txn_id: Optional[int] = None,
-    timeout: float = 300.0
+    timeout: float = TRANSACTION_TIMEOUT
 ):
     """Start internal CMAPI transaction.
 
@@ -78,7 +85,7 @@ def start_transaction(
     :param txn_id: id for transaction to start, defaults to None
     :type txn_id: Optional[int], optional
     :param timeout: time in seconds for cmapi transaction lock before it ends
-                    automatically, defaults to 300
+                    automatically, defaults to TRANSACTION_TIMEOUT
     :type timeout: float, optional
     :return: (success, txn_id, nodes)
     :rtype: tuple[bool, int, list[str]]
@@ -315,8 +322,7 @@ def broadcast_new_config(
                                defaults to DEFAULT_SM_CONF_PATH
     :param test_mode: for test purposes, defaults to False TODO: remove
     :param nodes: nodes list for config put, defaults to None
-    :param timeout: timeout passing to gracefully stop DMLProc TODO: for next
+    :param timeout: timeout passing to gracefully stop DMLProc process,
-                    releases. Could affect all logic of broadcacting new config
     :param distribute_secrets: flag to distribute secrets to nodes
     :param stateful_config_dict: stateful config update dict to distribute to nodes
     :raises CMAPIBasicError: If Broadcasting config to nodes failed with errors
@@ -332,7 +338,7 @@ def broadcast_new_config(
     headers = {'x-api-key': key}
     if stateful_config_dict:
         body = {
-            'timeout': 300,
+            'timeout': DMLPROC_SHUTDOWN_TIMEOUT if timeout is None else timeout,
             'stateful_config_dict': stateful_config_dict,
             'only_stateful_config': True,
         }
@@ -348,7 +354,7 @@ def broadcast_new_config(
         body = {
             'manager': root.find('./ClusterManager').text,
             'revision': root.find('./ConfigRevision').text,
-            'timeout': 300,
+            'timeout': DMLPROC_SHUTDOWN_TIMEOUT if timeout is None else timeout,
             'config': config_text,
             'mcs_config_filename': cs_config_filename,
             'sm_config_filename': sm_config_filename,
@@ -386,7 +392,7 @@ def broadcast_new_config(
         async with create_traced_async_session() as session:
             try:
                 async with session.put(
-                    url, headers=headers, json=body, ssl=False, timeout=120
+                    url, headers=headers, json=body, ssl=False, timeout=LONG_REQUEST_TIMEOUT
                 ) as response:
                     resp_json =  await response.json(encoding='utf-8')
                     response.raise_for_status()

cmapi/cmapi_server/managers/process.py

@@ -3,11 +3,18 @@ from __future__ import annotations
 import logging
 import os.path
 import socket
+import time
 from time import sleep
 
 import psutil
 
-from cmapi_server.constants import ALL_MCS_PROGS, MCS_INSTALL_BIN, MCSProgs, ProgInfo
+from cmapi_server.constants import (
+    ALL_MCS_PROGS,
+    DMLPROC_SHUTDOWN_TIMEOUT,
+    MCS_INSTALL_BIN,
+    MCSProgs,
+    ProgInfo,
+)
 from cmapi_server.exceptions import CMAPIBasicError
 from cmapi_server.process_dispatchers.base import BaseDispatcher
 from cmapi_server.process_dispatchers.container import ContainerDispatcher
@@ -238,32 +245,49 @@ class MCSProcessManager:
         return True
 
     @classmethod
-    def _wait_for_DMLProc_stop(cls, timeout: int = 10) -> bool:
+    def _wait_for_DMLProc_stop(cls, timeout: int = DMLPROC_SHUTDOWN_TIMEOUT) -> bool:
         """Waiting DMLProc process to stop.
 
-        :param timeout: timeout to wait, defaults to 10
+        :param timeout: timeout to wait in seconds, defaults to DMLPROC_SHUTDOWN_TIMEOUT
         :type timeout: int, optional
         :return: True on success
         :rtype: bool
         """
         logging.info(f'Waiting for DMLProc to stop in {timeout} seconds')
-        dmlproc_stopped = False
+        # Use a deadline-based loop with throttled logging to reduce noise.
-        while timeout > 0:
+        deadline = time.monotonic() + max(1, int(timeout))
-            logging.info(
+        LOG_INTERVAL = 30  # seconds
-                f'Waiting for DMLProc to stop. Seconds left {timeout}.'
+        next_log_in = 0  # log immediately on first iteration
-            )
+
+        while True:
+            remaining = int(deadline - time.monotonic())
+            if remaining <= 0:
+                break
+
             if not Process.check_process_alive('DMLProc'):
                 logging.info('DMLProc gracefully stopped by DBRM command.')
-                dmlproc_stopped = True
+                return True
-                break
+
-            sleep(1)
+            # Throttle waiting logs to roughly once every LOG_INTERVAL seconds
-            timeout -= 1
+            if next_log_in <= 0:
-        else:
+                sleep_for = min(10, remaining)
-            logging.error(
+                logging.info(
-                f'DMLProc did not stopped gracefully by DBRM command within '
+                    (
-                f'{timeout} seconds. Will be stopped directly.'
+                        f'Waiting for DMLProc to stop. Seconds left ~{remaining}. '
-            )
+                        f'Sleeping {sleep_for} seconds before next check.'
-        return dmlproc_stopped
+                    )
+                )
+                next_log_in = LOG_INTERVAL
+
+            sleep_for = min(10, remaining)
+            sleep(sleep_for)
+            next_log_in -= sleep_for
+
+        logging.error(
+            'DMLProc didn\'t stop gracefully by DBRM command within '
+            f'{timeout} seconds. Will be stopped directly.'
+        )
+        return False
 
     @classmethod
     def noop(cls, *args, **kwargs):
@@ -324,7 +348,7 @@ class MCSProcessManager:
 
     @classmethod
     def stop(
-        cls, name: str, is_primary: bool, use_sudo: bool, timeout: int = 10
+        cls, name: str, is_primary: bool, use_sudo: bool, timeout: int = DMLPROC_SHUTDOWN_TIMEOUT
     ) -> bool:
         """Stop mcs process.
 
@@ -455,7 +479,7 @@ class MCSProcessManager:
         cls,
         is_primary: bool,
         use_sudo: bool = True,
-        timeout: int = 10,
+        timeout: int = DMLPROC_SHUTDOWN_TIMEOUT,
     ):
         """Stop mcs node processes.
 
@@ -472,7 +496,7 @@ class MCSProcessManager:
         # undefined behaviour when primary gone and then recovers (failover
         # triggered 2 times).
         for prog_name in cls._get_sorted_progs(is_primary=True, reverse=True):
-            if not cls.stop(prog_name, is_primary, use_sudo):
+            if not cls.stop(prog_name, is_primary, use_sudo, timeout=timeout):
                 logging.error(f'Process "{prog_name}" not stopped properly.')
                 raise CMAPIBasicError(f'Error while stopping "{prog_name}"')
 

cmapi/cmapi_server/managers/upgrade/repo.py

@@ -154,7 +154,6 @@ class MariaDBESRepoManager:
         :raises CMAPIBasicError: no latest version matched with latest tested
         :raises CMAPIBasicError: if request error
         :return: latest MDB version matched with latest tested major
-        :rtype: str
         """
         try:
             # Download the keyring file
@@ -174,7 +173,7 @@ class MariaDBESRepoManager:
             )
             latest_version_num = sorted(latest_version_nums, reverse=True)[0]
             logging.debug(
-                'Succesfully got latest MBD version number: '
+                'Succesfully got latest MDB version number: '
                 f'{latest_version_num}'
             )
         except requests.RequestException as exc:

cmapi/mcs_cluster_tool/cluster_app.py

@@ -159,6 +159,7 @@ def stop(
         #       could affect put_config (helpers.py broadcast_config) operation
         timeout = 0
 
+    #TODO: bypass timeout here
     resp = client.shutdown_cluster({'in_transaction': True})
     return {'timestamp': start_time}
 

cmapi/mcs_cluster_tool/tools_commands.py

@@ -19,8 +19,12 @@ from rich.table import Table
 
 
 from cmapi_server.constants import (
-    MCS_DATA_PATH, MCS_SECRETS_FILENAME, REQUEST_TIMEOUT, TRANSACTION_TIMEOUT,
+    CMAPI_CONF_PATH,
-    CMAPI_CONF_PATH, CMAPI_PORT,
+    CMAPI_PORT,
+    MCS_DATA_PATH,
+    MCS_SECRETS_FILENAME,
+    REQUEST_TIMEOUT,
+    TRANSACTION_TIMEOUT,
 )
 from cmapi_server.controllers.api_clients import (
     AppControllerClient, ClusterControllerClient, NodeControllerClient