feature: Linux 2.0 Agent / Proxy 支持配置进程托管模式 (closed TencentBlueKing#1698

)

apps/backend/agent/artifact_builder/base.py

@@ -22,7 +22,7 @@
 from apps.backend import exceptions
 from apps.core.files import core_files_constants
 from apps.core.files.storage import get_storage
-from apps.node_man import constants
+from apps.node_man import constants, models
 from apps.utils import cache, files
 
 logger = logging.getLogger("app")
@@ -211,7 +211,21 @@ def _inject_dependencies(self, extract_dir: str):
             with self.storage.open(gsectl_file_path, mode="rb") as fs:
                 # mode 指定 w，覆盖现有文件
                 with open(gsectl_target_file_path, mode="wb") as local_fs:
-                    local_fs.write(fs.read())
+                    gsectl_file_content: str = fs.read().decode()
+                    # TODO 后续如果 1.0 Agent 包也纳入管理，此处需要区分
+                    #  1.0 Agent / Proxy 的 gsectl 采用固定的 rclocal 模式
+                    if gsectl_filename == "gsectl" and "{{ AUTO_TYPE }}" in gsectl_file_content:
+                        auto_type: str = models.GlobalSettings.get_config(
+                            models.GlobalSettings.KeyEnum.GSE2_LINUX_AUTO_TYPE.value,
+                            constants.GseLinuxAutoType.RCLOCAL.value,
+                        )
+                        logger.info(f"apply auto_type -> {auto_type} to gsectl")
+                        gsectl_file_content = gsectl_file_content.replace("{{ AUTO_TYPE }}", auto_type)
+                    if "{{ AUTO_TYPE }}" in gsectl_file_content:
+                        raise exceptions.PkgMetaInfoValidationError(
+                            _("渲染 AUTO_TYPE -> {auto_type} 到 gsectl 失败").format(auto_type=auto_type)
+                        )
+                    local_fs.write(gsectl_file_content.encode(encoding="utf-8"))
             logger.info(f"copy gsectl -> {gsectl_file_path} to {gsectl_target_file_path} success.")
 
             # 为二进制文件授予可执行权限

apps/backend/components/collections/agent_new/reload_agent_config.py

@@ -39,8 +39,12 @@ def get_script_content(self, data, common_data: AgentCommonData, host: models.Ho
         general_node_type = self.get_general_node_type(host.node_type)
         setup_path = common_data.host_id__ap_map[host.bk_host_id].get_agent_config(host.os_type)["setup_path"]
         agent_path = path_handler.join(setup_path, general_node_type, "bin")
-
-        return f"cd {agent_path} && ./gse_agent --reload"
+        if common_data.agent_step_adapter.is_legacy:
+            return f"cd {agent_path} && ./gse_agent --reload"
+        else:
+            # 趋势：gsectl 后续作为 Agent 操作的入口
+            # || 兼容旧版本 gsectl 没有 reload 操作
+            return f"cd {agent_path} && ./gsectl --reload agent || ./gse_agent --reload"
 
     def update_host_upstream_nodes(self, common_data: AgentCommonData, gse_version: str):
         hosts = [host for host in common_data.host_id_obj_map.values() if host.bk_cloud_id != constants.DEFAULT_CLOUD]

apps/backend/components/collections/agent_new/run_upgrade_command.py

@@ -13,7 +13,6 @@
 
 from django.conf import settings
 
-from apps.backend.agent.artifact_builder.proxy import ProxyArtifactBuilder
 from apps.node_man import constants, models
 
 from .base import AgentCommonData, AgentExecuteScriptService
@@ -50,13 +49,27 @@
 # Agent 重载配置命令模板
 AGENT_RELOAD_CMD_TEMPLATE = "cd {setup_path}/{node_type}/bin && ./{procs} --reload || ./gsectl restart all"
 
+
+# 进程拉起配置命令
+PROCESS_PULL_CONFIGURATION_CMD = """setup_startup_scripts
+remove_crontab
+"""
+
 # 节点类型 - 重载命令模板映射关系
-NODE_TYPE__RELOAD_CMD_TPL_MAP = {
+LEGACY_NODE_TYPE__RELOAD_CMD_TPL_MAP = {
     constants.NodeType.PROXY.lower(): PROXY_RELOAD_CMD_TEMPLATE,
     constants.NodeType.AGENT.lower(): AGENT_RELOAD_CMD_TEMPLATE,
 }
 
 
+# 新版本 Agent 具有 systemd / crontab 等多种拉起方式，systemd 模式下 reload 会导致进程拉不起来
+# 鉴于 reload 和 start 逻辑均是采取干掉进程再恢复的逻辑，统一采用 restart
+NODE_TYPE__RELOAD_CMD_TPL_MAP = {
+    constants.NodeType.PROXY.lower(): "cd {setup_path}/{node_type}/bin && ./gsectl restart all",
+    constants.NodeType.AGENT.lower(): "cd {setup_path}/{node_type}/bin && ./gsectl restart",
+}
+
+
 class RunUpgradeCommandService(AgentExecuteScriptService):
     @property
     def script_name(self):
@@ -79,23 +92,32 @@ def get_script_content(self, data, common_data: AgentCommonData, host: models.Ho
             with open(tpl_path, encoding="utf-8") as fh:
                 scripts = fh.read()
 
-            if host.node_type == constants.NodeType.PROXY:
-                if common_data.agent_step_adapter.is_legacy:
+            if common_data.agent_step_adapter.is_legacy:
+                process_pull_configuration_cmd: str = PROCESS_PULL_CONFIGURATION_CMD
+
+                if host.node_type == constants.NodeType.PROXY:
                     procs: typing.List[str] = ["gse_agent", "gse_transit", "gse_btsvr", "gse_data"]
                 else:
-                    procs: typing.List[str] = ["gse_agent"] + ProxyArtifactBuilder.PROXY_SVR_EXES
+                    procs: typing.List[str] = ["gse_agent"]
+
+                reload_cmd = LEGACY_NODE_TYPE__RELOAD_CMD_TPL_MAP[general_node_type].format(
+                    setup_path=agent_config["setup_path"], node_type=general_node_type, procs=" ".join(procs)
+                )
+
             else:
-                procs: typing.List[str] = ["gse_agent"]
+                # 新版本 Agent，通过 gsectl 配置进程拉起方式
+                process_pull_configuration_cmd: str = ""
+                reload_cmd = NODE_TYPE__RELOAD_CMD_TPL_MAP[general_node_type].format(
+                    setup_path=agent_config["setup_path"], node_type=general_node_type
+                )
 
-            reload_cmd = NODE_TYPE__RELOAD_CMD_TPL_MAP[general_node_type].format(
-                setup_path=agent_config["setup_path"], node_type=general_node_type, procs=" ".join(procs)
-            )
             scripts = scripts.format(
                 setup_path=agent_config["setup_path"],
                 temp_path=agent_config["temp_path"],
                 package_name=agent_upgrade_pkg_name,
                 node_type=general_node_type,
                 reload_cmd=reload_cmd,
+                process_pull_configuration_cmd=process_pull_configuration_cmd,
                 pkg_cpu_arch=host.cpu_arch,
             )
             return scripts

apps/backend/exceptions.py

@@ -83,3 +83,8 @@ class AgentConfigTemplateNotExistError(BackendBaseException):
 class NotSemanticVersionError(BackendBaseException):
     MESSAGE_TPL = _("版本号 -> {version} 不符合语义化版本规则")
     ERROR_CODE = 13
+
+
+class PkgMetaInfoValidationError(BackendBaseException):
+    MESSAGE = _("安装包元数据校验失败")
+    ERROR_CODE = 14

apps/backend/tests/components/collections/agent_new/test_run_upgrade_command.py

@@ -18,6 +18,8 @@
 )
 from apps.backend.components.collections.agent_new.run_upgrade_command import (
     AGENT_RELOAD_CMD_TEMPLATE,
+    NODE_TYPE__RELOAD_CMD_TPL_MAP,
+    PROCESS_PULL_CONFIGURATION_CMD,
     WINDOWS_UPGRADE_CMD_TEMPLATE,
 )
 from apps.node_man import constants, models
@@ -39,6 +41,7 @@ class RunUpgradeCommandSuccessTestCase(base.JobBaseTestCase):
         package_name="gse_client-linux-x86_64_upgrade.tgz",
         node_type="agent",
         reload_cmd=AGENT_RELOAD_CMD_TEMPLATE.format(setup_path="/usr/local/gse", node_type="agent", procs="gse_agent"),
+        process_pull_configuration_cmd=PROCESS_PULL_CONFIGURATION_CMD,
         pkg_cpu_arch="x86_64",
     )
     WINDOWS_TEST_SCRIPTS = WINDOWS_UPGRADE_CMD_TEMPLATE.format(
@@ -93,9 +96,8 @@ def tearDown(self) -> None:
             temp_path="/tmp",
             package_name="gse_agent-2.0.0.tgz",
             node_type="agent",
-            reload_cmd=AGENT_RELOAD_CMD_TEMPLATE.format(
-                setup_path="/usr/local/gse", node_type="agent", procs="gse_agent"
-            ),
+            reload_cmd=NODE_TYPE__RELOAD_CMD_TPL_MAP["agent"].format(setup_path="/usr/local/gse", node_type="agent"),
+            process_pull_configuration_cmd="",
             pkg_cpu_arch="x86_64",
         )
         for record_result in record[JobApi.fast_execute_script]:

apps/node_man/constants.py

@@ -965,6 +965,22 @@ def _get_member__alias_map(cls) -> Dict[Enum, str]:
         }
 
 
+class GseLinuxAutoType(EnhanceEnum):
+    """GSE Linux 进程拉起（托管方式），该配置在 Agent 打包时，渲染到 gsectl 中"""
+
+    CRONTAB = "crontab"
+    RCLOCAL = "rclocal"
+    SYSTEMD = "systemd"
+
+    @classmethod
+    def _get_member__alias_map(cls) -> Dict[Enum, str]:
+        return {
+            cls.CRONTAB: _("crontab"),
+            cls.RCLOCAL: _("rclocal"),
+            cls.SYSTEMD: _("systemd"),
+        }
+
+
 ########################################################################################################
 # CMDB
 ########################################################################################################

apps/node_man/models.py

@@ -115,6 +115,8 @@ class KeyEnum(Enum):
         PLUGIN_COMMON_CONSTANTS = "PLUGIN_COMMON_CONSTANTS"
         # GSE 2.0 灰度列表
         GSE2_GRAY_SCOPE_LIST = "GSE2_GRAY_SCOPE_LIST"
+        # GSE 2.0 Linux Agent/Proxy 进程托管（拉起）方式，默认为 rclocal
+        GSE2_LINUX_AUTO_TYPE = "GSE2_LINUX_AUTO_TYPE"
         # 禁用订阅业务列表
         DISABLE_SUBSCRIPTION_SCOPE_LIST = "DISABLE_SUBSCRIPTION_SCOPE_LIST"
         GSE2_GRAY_AP_MAP = "GSE2_GRAY_AP_MAP"

script_tools/agent_tools/agent2/setup_agent.sh

@@ -864,13 +864,14 @@ DEBUG_LOG_FILE=${TMP_DIR}/nm.${0##*/}.${TASK_ID}.debug
 exec &> >(tee "$DEBUG_LOG_FILE")
 
 log check_env - "Args are: $*"
+
+# removed remove_crontab、setup_startup_scripts -> 由 gsectl 判断是否添加 / 移除
+
 for step in check_env \
             download_pkg \
-            remove_crontab \
             remove_agent \
             remove_proxy_if_exists \
             setup_agent \
-            setup_startup_scripts \
             check_deploy_result; do
     $step
 done

script_tools/agent_tools/agent2/setup_proxy.sh

@@ -807,11 +807,9 @@ log check_env - "$@"
 #pre_view
 for step in check_env \
             download_pkg \
-            remove_crontab \
             remove_proxy \
             remove_agent_if_exists \
             setup_proxy \
-            setup_startup_scripts \
             check_deploy_result; do
     $step
 done