metallb: Speaker does not respon ARP requests

[BUG]: Speaker does not respon ARP request - Version of MetalLB : 0.7.3 - Version of Kubernetes: 1.13.3 - Network addon: WeaveNet 2.5.1 - Kube-proxy using iptables

I got many services with different IPs:

Speaker not responded ARP to 10.30.225.3

[root@bdprden0003]# arping -I eno49 10.30.225.2
ARPING 10.30.225.2 from 10.30.225.37 eno49
Unicast reply from 10.30.225.2 [48:DF:37:00:FF:B8]  0.869ms
Unicast reply from 10.30.225.2 [48:DF:37:00:FF:B8]  0.755ms
Unicast reply from 10.30.225.2 [48:DF:37:00:FF:B8]  0.852ms
Unicast reply from 10.30.225.2 [48:DF:37:00:FF:B8]  0.797ms
^CSent 4 probes (1 broadcast(s))
Received 4 response(s)

[root@bdprden0003]# arping -I eno49 10.30.225.4
ARPING 10.30.225.4 from 10.30.225.37 eno49
Unicast reply from 10.30.225.4 [48:DF:37:00:FF:B8]  1.117ms
Unicast reply from 10.30.225.4 [48:DF:37:00:FF:B8]  0.899ms
Unicast reply from 10.30.225.4 [48:DF:37:00:FF:B8]  1.219ms
Unicast reply from 10.30.225.4 [48:DF:37:00:FF:B8]  0.861ms
Unicast reply from 10.30.225.4 [48:DF:37:00:FF:B8]  0.924ms
Unicast reply from 10.30.225.4 [48:DF:37:00:FF:B8]  0.856ms
Unicast reply from 10.30.225.4 [48:DF:37:00:FF:B8]  0.811ms
^CSent 7 probes (1 broadcast(s))
Received 7 response(s)

[root@bdprden0003]# arping -I eno49 10.30.225.3
ARPING 10.30.225.3 from 10.30.225.37 eno49
^CSent 96 probes (96 broadcast(s))
Received 0 response(s)
~ $ kubectl get services --all-namespaces | grep 10.30.225.2
kube-cep      cep-api-deploy-http                        LoadBalancer   10.100.223.168   10.30.225.2   8100:30234/TCP                        21d
kube-cep      cep-api-uat-http                           LoadBalancer   10.107.40.58     10.30.225.2   8104:31615/TCP                        59d
kube-cep      cep-mysql-public                           LoadBalancer   10.105.168.255   10.30.225.2   3306:30524/TCP                        81d
kube-cep      cep-ui-http                                LoadBalancer   10.108.138.36    10.30.225.2   80:32289/TCP                          92d
kube-cep      cep-ui-ssh                                 LoadBalancer   10.96.184.145    10.30.225.2   9022:31580/TCP                        87d
kube-cep      cep-ui-uat-http                            LoadBalancer   10.109.155.118   10.30.225.2   8004:32356/TCP                        79d

~ $ kubectl get services --all-namespaces | grep 10.30.225.3
kong          kong-admin                                 LoadBalancer   10.98.22.247     10.30.225.3   8001:30911/TCP                        88d
kong          kong-proxy                                 LoadBalancer   10.104.102.126   10.30.225.3   8000:30896/TCP                        88d
kong          konga-web-public                           LoadBalancer   10.107.108.97    10.30.225.3   1337:30890/TCP                        86d
kube-db       postgres96-public                          LoadBalancer   10.104.71.180    10.30.225.3   5432:31599/TCP                        90d
prometheus    grafana                                    LoadBalancer   10.97.59.67      10.30.225.3   3000:32451/TCP                        86d
zabbix        zabbix-web-public                          LoadBalancer   10.103.114.17    10.30.225.3   8080:30779/TCP                        86d

~ $ kubectl get services --all-namespaces | grep 10.30.225.4
cdh-prod      pg96-cdh-prod-public                       LoadBalancer   10.105.231.194   10.30.225.4   5432:32571/TCP                        57d

Any help?

About this issue

  • Original URL
  • State: closed
  • Created 5 years ago
  • Reactions: 3
  • Comments: 24 (8 by maintainers)

Most upvoted comments

I’m also having this issue, it only happens on one of my nodes and for one of my services. I can’t see any obvious difference between the working services and the one that doesnt work, but killing the speaker each time it happens reliably fixes it until the pod is restarted.