Después de que Kubernetes construyó un clúster a través del componente de red de franela, no pudieron acceder entre sí a través del segmento de red de franela y se agotó el tiempo de espera del ping del nodo maestro 10.244.2.0 (segmento de red de franela del nodo 1).
kubernetes versión 1.24.2
Información de la tarjeta de red virtual master franela
[root@master1 k8s]# ifconfig flannel.1
flannel.1 Link encap:Ethernet HWaddr 56:94:4B:2E:EA:5D
inet addr:10.244.0.0 Bcast:0.0.0.0 Mask:255.255.255.255
inet6 addr: fe80::5494:4bff:fe2e:ea5d/64 Scope:Link
UP BROADCAST RUNNING MULTICAST MTU:1450 Metric:1
RX packets:0 errors:0 dropped:0 overruns:0 frame:0
TX packets:988 errors:0 dropped:8 overruns:0 carrier:0
collisions:0 txqueuelen:0
RX bytes:0 (0.0 B) TX bytes:58486 (57.1 KiB)
Los registros de solicitudes capturados por tcpdump son los siguientes
[root@master1 k8s]# tcpdump -i eth0 port 30000
tcpdump: verbose output suppressed, use -v or -vv for full protocol decode
listening on eth0, link-type EN10MB (Ethernet), capture size 262144 bytes
20:47:33.186133 IP 172.16.29.34.60629 > master1.ndmps: Flags [S], seq 902841225, win 64240, options [mss 1460,nop,wscale 8,nop,nop,sackOK], length 0
20:47:33.203308 IP 172.16.29.34.60630 > master1.ndmps: Flags [S], seq 2938533821, win 64240, options [mss 1460,nop,wscale 8,nop,nop,sackOK], length 0
20:47:33.438264 IP 172.16.29.34.60633 > master1.ndmps: Flags [S], seq 2972658478, win 64240, options [mss 1460,nop,wscale 8,nop,nop,sackOK], length 0
20:47:34.200834 IP 172.16.29.34.60629 > master1.ndmps: Flags [S], seq 902841225, win 64240, options [mss 1460,nop,wscale 8,nop,nop,sackOK], length 0
20:47:34.215896 IP 172.16.29.34.60630 > master1.ndmps: Flags [S], seq 2938533821, win 64240, options [mss 1460,nop,wscale 8,nop,nop,sackOK], length 0
20:47:34.447110 IP 172.16.29.34.60633 > master1.ndmps: Flags [S], seq 2972658478, win 64240, options [mss 1460,nop,wscale 8,nop,nop,sackOK], length 0
20:47:36.214314 IP 172.16.29.34.60629 > master1.ndmps: Flags [S], seq 902841225, win 64240, options [mss 1460,nop,wscale 8,nop,nop,sackOK], length 0
[root@master1 ~]# tcpdump -i flannel.1
tcpdump: verbose output suppressed, use -v or -vv for full protocol decode
listening on flannel.1, link-type EN10MB (Ethernet), capture size 262144 bytes
20:47:33.186218 IP master1.63168 > 10.244.2.3.pcsync-https: Flags [S], seq 902841225, win 64240, options [mss 1460,nop,wscale 8,nop,nop,sackOK], length 0
20:47:33.203355 IP master1.47411 > 10.244.2.3.pcsync-https: Flags [S], seq 2938533821, win 64240, options [mss 1460,nop,wscale 8,nop,nop,sackOK], length 0
20:47:33.438341 IP master1.37028 > 10.244.2.3.pcsync-https: Flags [S], seq 2972658478, win 64240, options [mss 1460,nop,wscale 8,nop,nop,sackOK], length 0
20:47:34.200885 IP master1.63168 > 10.244.2.3.pcsync-https: Flags [S], seq 902841225, win 64240, options [mss 1460,nop,wscale 8,nop,nop,sackOK], length 0
20:47:34.215930 IP master1.47411 > 10.244.2.3.pcsync-https: Flags [S], seq 2938533821, win 64240, options [mss 1460,nop,wscale 8,nop,nop,sackOK], length 0
20:47:34.447159 IP master1.37028 > 10.244.2.3.pcsync-https: Flags [S], seq 2972658478, win 64240, options [mss 1460,nop,wscale 8,nop,nop,sackOK], length 0
Información de ruta
[root@master1 k8s]# ip route
default via 172.16.103.254 dev eth0 proto static metric 100
10.244.1.0/24 via 10.244.1.0 dev flannel.1 onlink
10.244.2.0/24 via 10.244.2.0 dev flannel.1 onlink
172.16.103.0/24 dev eth0 proto kernel scope link src 172.16.103.66 metric 100
172.17.0.0/16 dev docker0 proto kernel scope link src 172.17.0.1
Información ARP y puente
[root@master1 k8s]# arp -n
? (10.244.2.0) at 3e:18:6e:82:70:79 [ether] PERM on flannel.1
? (172.16.103.253) at 08:4f:0a:70:6d:99 [ether] on eth0
? (172.16.103.251) at 14:96:2d:4d:3d:0f [ether] on eth0
? (10.244.1.0) at 0a:31:9f:e4:b3:14 [ether] PERM on flannel.1
? (172.16.103.68) at fe:fc:fe:ff:13:65 [ether] on eth0
? (172.16.103.67) at fe:fc:fe:ff:ee:0c [ether] on eth0
? (172.16.103.254) at 34:00:a3:3f:26:f9 [ether] on eth0
[root@master1 k8s]# bridge fdb | grep 3e:18:6e:82:70:79
3e:18:6e:82:70:79 dev flannel.1 dst 172.16.103.67 self permanent
Y el nodo maestro puede acceder a la dirección IP real del nodo (172.16.103.67) (lo que debería descartar problemas con el grupo de seguridad y el firewall), sin mensajes de error para el canal kube, el proxy kube y los coredns.
¿Cómo debemos solucionar o encontrar la causa?