Redis Cluster - Slave 不 FAILOVER

Question

我已经创建了一个 Redis 集群，如下所示。

xxx.xxx.xxx.195:9100    xxx.xxx.xxx.196:9100    xxx.xxx.xxx.197:9100
xxx.xxx.xxx.195:9200    xxx.xxx.xxx.196:9200    xxx.xxx.xxx.197:9200

我遇到过，当我同时停止作为集群主节点的 2 个 Redis 实例(xxx.xxx.xxx.196 的 2 个实例)时，集群无法恢复，

xxx.xxx.xxx.195:9100 (Master)    xxx.xxx.xxx.196:9100 (Master)   xxx.xxx.xxx.197:9100 (Slave)
xxx.xxx.xxx.195:9200 (Slave)     xxx.xxx.xxx.196:9200 (Master)   xxx.xxx.xxx.197:9200 (Slave)

但同时，如果我停止 2 个 .195 服务器实例，其中 9100 -Master 和 9200 - Slave。集群恢复正常

集群配置文件:

protected-mode no
activerehashing yes
cluster-enabled yes
cluster-config-file /opt/redis/conf/nodes9100.conf
cluster-slave-validity-factor 0
cluster-node-timeout 5000
appendonly yes

Redis登录专用从服务器:

28939:S 09 Oct 16:08:32.834 - 0 clients connected (0 slaves), 1327200 bytes in use
28939:S 09 Oct 16:08:32.834 * Connecting to MASTER xxx.xxx.xxx.196:9200
28939:S 09 Oct 16:08:32.835 * MASTER <-> SLAVE sync started
28939:S 09 Oct 16:08:32.835 # Error condition on socket for SYNC: Connection refused
28939:S 09 Oct 16:08:33.837 * Connecting to MASTER xxx.xxx.xxx.196:9200
28939:S 09 Oct 16:08:33.837 * MASTER <-> SLAVE sync started
28939:S 09 Oct 16:08:33.837 # Error condition on socket for SYNC: Connection refused
28939:S 09 Oct 16:08:34.839 * Connecting to MASTER xxx.xxx.xxx.196:9200
28939:S 09 Oct 16:08:34.839 * MASTER <-> SLAVE sync started
28939:S 09 Oct 16:08:34.839 # Error condition on socket for SYNC: Connection refused
28939:S 09 Oct 16:08:35.840 * Connecting to MASTER xxx.xxx.xxx.196:9200
28939:S 09 Oct 16:08:35.840 * MASTER <-> SLAVE sync started
28939:S 09 Oct 16:08:35.840 # Error condition on socket for SYNC: Connection refused
28939:S 09 Oct 16:08:36.744 - Node 982d9b0a50b393d5fe604caefc0acaae68547648 reported node b57d59fb5685daeaac7e249d99fa257e9be66f4f as not reachable.
28939:S 09 Oct 16:08:36.844 * Connecting to MASTER xxx.xxx.xxx.196:9200
28939:S 09 Oct 16:08:36.844 * MASTER <-> SLAVE sync started
28939:S 09 Oct 16:08:36.844 # Error condition on socket for SYNC: Connection refused

Answer 1

发现问题:
来源:https://redis.io/topics/cluster-tutorialRedis 集群还在分区期间提供一定程度的可用性，即在实际情况下，当某些节点出现故障或无法通信时继续操作的能力。但是，如果发生更大的故障(例如，当大多数主节点不可用时)，集群将停止运行。