fix(redis): 自愈集群信息修复 #2592

TencentBlueKing · Dec 15, 2023 · 31d0917 · 31d0917
1 parent d44658b
commit 31d0917
Show file tree

Hide file tree

Showing 3 changed files with 49 additions and 12 deletions.
diff --git a/dbm-ui/backend/db_meta/api/cluster/apis.py b/dbm-ui/backend/db_meta/api/cluster/apis.py
@@ -116,4 +116,4 @@ def query_cluster_by_hosts(hosts: List):
 
     for machine in machine_cluster.values():
         machine["ports"] = machine_ports.get(machine["ip"])
-    return machine_cluster.values()
+    return list(machine_cluster.values())
diff --git a/dbm-ui/backend/db_services/redis/autofix/bill.py b/dbm-ui/backend/db_services/redis/autofix/bill.py
@@ -78,7 +78,7 @@ def create_ticket(cluster: RedisAutofixCore, redis_proxies: list, redis_slaves:
         ticket_type=TicketType.REDIS_CLUSTER_AUTOFIX.value,
         group=DBType.Redis.value,
         status=TicketStatus.PENDING.value,
-        remark=_("自动发起-自愈任务-{cluster.immute_domain}"),
+        remark=_("自动发起-自愈任务-{}".format(cluster.immute_domain)),
         details=details,
         is_reviewed=True,
     )

diff --git a/dbm-ui/backend/ticket/builders/redis/redis_toolbox_autofix.py b/dbm-ui/backend/ticket/builders/redis/redis_toolbox_autofix.py
@@ -13,6 +13,8 @@
 from rest_framework import serializers
 
 from backend.configuration.constants import AffinityEnum
+from backend.db_meta.enums import InstanceRole
+from backend.db_meta.models import Cluster, StorageInstance
 from backend.db_services.dbbase.constants import IpSource
 from backend.flow.engine.controller.redis import RedisController
 from backend.ticket import builders
@@ -52,12 +54,16 @@ def post_callback(self):
         ticket_data = next_flow.details["ticket_data"]
 
         for info_index, info in enumerate(self.ticket_data["infos"]):
-            for role in ["redis_master", "proxy", "redis_slave"]:
+            for role in [
+                InstanceRole.REDIS_MASTER.value,
+                InstanceRole.REDIS_PROXY.value,
+                InstanceRole.REDIS_SLAVE.value,
+            ]:
                 role_hosts = info.get(role)
                 if not role_hosts:
                     continue
 
-                role_group = "backend_group" if role == "redis_master" else role
+                role_group = "backend_group" if role == InstanceRole.REDIS_MASTER.value else role
                 for role_host_index, role_host in enumerate(role_hosts):
                     role_host["target"] = nodes.get(f"{info_index}_{role_group}")[role_host_index]
 
@@ -88,18 +94,49 @@ def patch_ticket_detail(self):
         super().patch_ticket_detail()
 
         resource_spec = {}
+        cluster_ids = [infos["cluster_id"] for infos in self.ticket.details["infos"]]
+        id__cluster = {cluster.id: cluster for cluster in Cluster.objects.filter(id__in=cluster_ids)}
         for info in self.ticket.details["infos"]:
-            for role in ["redis_master", "proxy", "redis_slave"]:
+            cluster = id__cluster[info["cluster_id"]]
+            for role in [
+                InstanceRole.REDIS_MASTER.value,
+                InstanceRole.REDIS_PROXY.value,
+                InstanceRole.REDIS_SLAVE.value,
+            ]:
                 role_hosts = info.get(role)
+
                 if not role_hosts:
                     continue
-                role_group = "backend_group" if role == "redis_master" else role
-                role_group_affinity = AffinityEnum.CROS_SUBZONE if role_group == "backend_group" else AffinityEnum.NONE
-                resource_spec[role_group] = {
-                    "spec_id": info[role][0]["spec_id"],
-                    "count": len(role_hosts),
-                    "affinity": role_group_affinity.value,
-                }
+
+                if role == InstanceRole.REDIS_MASTER.value:
+                    # 如果替换角色是master，则是master/slave成对替换
+                    resource_spec["backend_group"] = {
+                        "spec_id": info[role][0]["spec_id"],
+                        "count": len(role_hosts),
+                        "location_spec": {"city": cluster.region, "sub_zone_ids": []},
+                        "affinity": cluster.disaster_tolerance_level,
+                    }
+                elif role == InstanceRole.REDIS_SLAVE.value:
+                    # 如果是替换slave， 需要和当前集群中的配对的 master 不同机房
+                    redis_slaves = StorageInstance.objects.prefetch_related("as_receiver", "machine").filter(
+                        cluster=cluster, machine__ip__in=[host["ip"] for host in role_hosts]
+                    )
+                    ip__redis_slave = {slave.machine.ip: slave for slave in redis_slaves}
+                    for role_host in role_hosts:
+                        redis_master = ip__redis_slave[role_host["ip"]].as_receiver.get().ejector
+                        resource_spec[f"{role}_{role_host['ip']}"] = {
+                            "spec_id": role_host["spec_id"],
+                            "count": 1,
+                            "location_spec": {
+                                "city": cluster.region,
+                                "sub_zone_ids": [redis_master.machine.bk_sub_zone_id],
+                                "include_or_exclue": False,
+                            },
+                        }
+                elif role == InstanceRole.REDIS_PROXY.value:
+                    # TODO: proxy替换的亲和性需要衡量存量proxy的分布，暂时忽略
+                    resource_spec[role] = {"spec_id": info[role][0]["spec_id"], "count": len(role_hosts)}
+
             info["resource_spec"] = resource_spec
 
         self.ticket.save(update_fields=["details"])