fix: 原地slave重建生成单据前检查实例是否存活 TencentBlueKing#8865

ygcyao · Jan 2, 2025 · e79a8c8 · e79a8c8
1 parent f08ed98
commit e79a8c8
Show file tree

Hide file tree

Showing 2 changed files with 32 additions and 0 deletions.
diff --git a/dbm-ui/backend/flow/engine/bamboo/scene/mysql/mysql_restore_slave_remote_flow.py b/dbm-ui/backend/flow/engine/bamboo/scene/mysql/mysql_restore_slave_remote_flow.py
@@ -15,9 +15,11 @@
 
 from django.utils.translation import ugettext as _
 
+from backend.components import DRSApi
 from backend.configuration.constants import DBType
 from backend.constants import IP_PORT_DIVIDER, IP_PORT_DIVIDER_FOR_DNS
 from backend.db_meta.enums import ClusterType, InstanceInnerRole, InstancePhase, InstanceStatus
+from backend.db_meta.exceptions import InstanceNotExistException
 from backend.db_meta.models import Cluster
 from backend.db_package.models import Package
 from backend.flow.consts import MediumEnum
@@ -433,6 +435,20 @@ def restore_local_slave_flow(self):
                 machine__ip=self.data["slave_ip"],
                 port=self.data["slave_port"],
             )
+
+            res = DRSApi.rpc(
+                {
+                    "addresses": [target_slave.ip_port],
+                    "cmds": ["select version()"],
+                    "force": False,
+                    "bk_cloud_id": target_slave.machine.bk_cloud_id,
+                }
+            )
+            if res[0]["error_msg"]:
+                raise InstanceNotExistException(
+                    _("请检查实例 {} 是否存活,是否正常可访问，slave原地重建是实例级别的，且必须保证实例存活方可提单进行").format(target_slave.ip_port)
+                )
+
             master = cluster_model.storageinstance_set.get(instance_inner_role=InstanceInnerRole.MASTER.value)
             self.data["new_slave_ip"] = target_slave.machine.ip
             self.data["bk_biz_id"] = cluster_model.bk_biz_id

diff --git a/dbm-ui/backend/flow/engine/bamboo/scene/spider/remote_local_slave_recover.py b/dbm-ui/backend/flow/engine/bamboo/scene/spider/remote_local_slave_recover.py
@@ -16,8 +16,10 @@
 from django.utils.crypto import get_random_string
 from django.utils.translation import ugettext as _
 
+from backend.components import DRSApi
 from backend.configuration.constants import DBType
 from backend.db_meta.enums import InstancePhase, InstanceStatus
+from backend.db_meta.exceptions import InstanceNotExistException
 from backend.db_meta.models import Cluster
 from backend.flow.engine.bamboo.scene.common.builder import Builder, SubBuilder
 from backend.flow.engine.bamboo.scene.common.get_file_list import GetFileList
@@ -115,6 +117,20 @@ def tendb_remote_slave_local_recover(self):
                 self.data["master_port"] = shard.storage_instance_tuple.ejector.port
                 self.data["slave_port"] = shard.storage_instance_tuple.receiver.port
                 target_slave = cluster_class.storageinstance_set.get(id=shard.storage_instance_tuple.receiver.id)
+                # 检查slave是否存活
+                res = DRSApi.rpc(
+                    {
+                        "addresses": [target_slave.ip_port],
+                        "cmds": ["select version()"],
+                        "force": False,
+                        "bk_cloud_id": target_slave.machine.bk_cloud_id,
+                    }
+                )
+                if res[0]["error_msg"]:
+                    raise InstanceNotExistException(
+                        _("请检查实例 {} 是否存活,是否正常可访问，slave原地重建是实例级别的，且必须保证实例存活方可提单进行").format(target_slave.ip_port)
+                    )
+
                 master = cluster_class.storageinstance_set.get(id=shard.storage_instance_tuple.ejector.id)
                 cluster = {
                     "phase": InstancePhase.TRANS_STAGE.value,