Kubernetes V1.19.13-kubeapi 服务器无法连接到不同的 etcd 数据库

Question

我已经将 Kubernets 集群（3 个主服务器、3 个 etcd 服务器数据库）从 升级1.18到v1.19.13，并将 ETCD 升级到3.4.13。由于 API 服务器不稳定，不断切换不同的 etcd 服务器，因为这个集群无法正常工作。这些集群在 CentOS steam 8 上运行。这个集群在升级之前可以正常工作，升级后我才看到这个问题。

有什么帮助可以解决这个问题吗？这个版本还有问题吗？

API 服务器日志

I0731 00:54:39.498953       1 client.go:360] parsed scheme: "passthrough"
I0731 00:54:39.499025       1 passthrough.go:48] ccResolverWrapper: sending update to cc: {[{https://0.0.0.02:2379  <nil> 0 <nil>}] <nil> <nil>}
I0731 00:54:39.499035       1 clientconn.go:948] ClientConn switching balancer to "pick_first"
I0731 00:54:40.241615       1 client.go:360] parsed scheme: "passthrough"
I0731 00:54:40.241681       1 passthrough.go:48] ccResolverWrapper: sending update to cc: {[{https://0.0.0.01:2379  <nil> 0 <nil>}] <nil> <nil>}
I0731 00:54:40.241691       1 clientconn.go:948] ClientConn switching balancer to "pick_first"
I0731 00:54:45.348969       1 client.go:360] parsed scheme: "passthrough"
I0731 00:54:45.349030       1 passthrough.go:48] ccResolverWrapper: sending update to cc: {[{https://0.0.0.03:2379  <nil> 0 <nil>}] <nil> <nil>}
I0731 00:54:45.349040       1 clientconn.go:948] ClientConn switching balancer to "pick_first"
I0731 00:55:16.460379       1 client.go:360] parsed scheme: "passthrough"
I0731 00:55:16.460428       1 passthrough.go:48] ccResolverWrapper: sending update to cc: {[{https://0.0.0.01:2379  <nil> 0 <nil>}] <nil> <nil>}
I0731 00:55:16.460439       1 clientconn.go:948] ClientConn switching balancer to "pick_first"
I0731 00:55:17.461906       1 client.go:360] parsed scheme: "passthrough"

etcd 看起来很健康

# /opt/bin/etcdctl.sh   version
etcdctl version: 3.4.13
API version: 3.4

 /opt/bin/etcdctl.sh  endpoint health
127.0.0.1:2379 is healthy: successfully committed proposal: took = 9.739533ms



# /opt/bin/etcdctl.sh  check perf
 60 / 60 Boooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooo! 100.00% 1m0s
PASS: Throughput is 150 writes/s
PASS: Slowest request took 0.042491s
PASS: Stddev is 0.001743s
PASS

# /opt/bin/etcdctl.sh  endpoint status --cluster -w table
+----------------------------+------------------+---------+---------+-----------+------------+-----------+------------+--------------------+--------+
|          ENDPOINT          |        ID        | VERSION | DB SIZE | IS LEADER | IS LEARNER | RAFT TERM | RAFT INDEX | RAFT APPLIED INDEX | ERRORS |
+----------------------------+------------------+---------+---------+-----------+------------+-----------+------------+--------------------+--------+
| https://0.0.0.02:2379 | 15cd65a732ebd5d8 |  3.4.13 |   26 MB |     false |      false |      9305 |   19813854 |           19813854 |        |
| https://0.0.0.03:2379 | add66a254676e690 |  3.4.13 |   26 MB |      true |      false |      9305 |   19813854 |           19813854 |        |
| https://0.0.0.01:2379 | e2811ed02ce71623 |  3.4.13 |   26 MB |     false |      false |      9305 |   19813854 |           19813854 |        |
+----------------------------+------------------+---------+---------+-----------+------------+-----------+------------+--------------------+--------+