k8s calico網路排錯2
解決了跨node的pod連通性，接下來面臨新問題，首先是nodeport只對pod所在node能夠訪問。
參考上面連結，執行sudo iptables -P FORWARD ACCEPT初步解決該問題，hadoop001的還是不行
報錯如下：
Chain KUBE-SERVICES (2 references)
target prot opt source destination
REJECT tcp -- anywhere anywhere /* pezy/hbase-hmaster:master-web has no endpoints */ ADDRTYPE match dst-type LOCAL tcp dpt:32098 reject-with icmp-port-unreachable
REJECT tcp -- anywhere 10.96.7.193 /* pezy/hbase-hmaster:master-web has no endpoints */ tcp dpt:60010 reject-with icmp-port-unreachable
REJECT tcp -- anywhere anywhere /* pezy/yarn-resourcemanager:web has no endpoints */ ADDRTYPE match dst-type LOCAL tcp dpt:30880 reject-with icmp-port-unreachable
REJECT tcp -- anywhere 10.97.228.2 /* pezy/yarn-resourcemanager:web has no endpoints */ tcp dpt:23188 reject-with icmp-port-unreachable
REJECT tcp -- anywhere anywhere /* pezy/yarn-resourcemanager:jobhistory-web has no endpoints */ ADDRTYPE match dst-type LOCAL tcp dpt:31879 reject-with icmp-port-unreachable
後發現hadoop001的docker映象，外部無法訪問，包括私有倉庫，重啟倉庫。
發現ip轉發沒有開。
[[email protected] ~]# docker run -d -p 5000:5000 -v /dlw_registry:/var/lib/registry registry:2.6.0
WARNING: IPv4 forwarding is disabled. Networking will not work.
c1e2695253639540d02f8bc4cbfc586d2a75b679b5275a6387e7aa5b45787d1f
解決辦法：
# vi /etc/sysctl.conf
或者
# vi /usr/lib/sysctl.d/00-system.conf
新增如下程式碼：
net.ipv4.ip_forward=1
重啟network服務
# systemctl restart network
檢視是否修改成功
# sysctl net.ipv4.ip_forward
如果返回為“net.ipv4.ip_forward = 1”則表示成功了
修改後，倉庫可以訪問，docker可以外部訪問。