Interpod Communication - kubernetes

We have the following in GoogleCloud Kubernetes:
3 REST API Pods which take POSTs and send them to clients that are connected via Websocket.
If one of those Pods is posted on, we want to send this post to all other pods.
The question is: How / Where can we find the IPs of the other Pods?

You can find the pod IP using the below and try hitting the IP but suggest you expose a service to do this.
kubectl get po -n test -o wide
NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES
pod1 1/1 Running 0 98m 10.42.0.16 worker3 <none> <none>
pod6 1/1 Running 0 87m 10.44.0.26 worker1 <none> <none>
To expose a service:
kubectl expose pod/pod1 -n test --name=svc --port=80 --target-port=80
kubectl run bb --image=busybox -n test -it --rm -- sh
If you don't see a command prompt, try pressing enter.
wget -O- svc:80
Connecting to svc:80 (10.101.174.245:80)
writing to stdout
<html><body><h1>It works!</h1></body></html>

Related

Kubespray : Netchecker connectivity check fails

I deployed a Kubernetes (v1.17.5) cluster on OpenStack instances using Kubespray. Those instances are CentOS 7.6.1811 qcow2 images imported in Glance.
The install was successful, and I can see my nodes and pods with kubectl commands.
I used the deploy_netchecker option to deploy NetChecker and test the network within my cluster, and set network_plugin="flannel".
I also tried kube_proxy_mode="iptables", but it doesn't seem to affect the result.
That's pretty much all the changes I did in the k8s-cluster.yml file.
All the pods are running, services too :
[centos#cl1-master-0 ~]$ kubectl get svc --all-namespaces
NAMESPACE NAME TYPE CLUSTER-IP EXTERNAL-IP PORT(S) AGE
default kubernetes ClusterIP 10.233.0.1 <none> 443/TCP 46h
default netchecker-service NodePort 10.233.13.213 <none> 8081:31081/TCP 46h
kube-system coredns ClusterIP 10.233.0.3 <none> 53/UDP,53/TCP,9153/TCP 46h
kube-system dashboard-metrics-scraper ClusterIP 10.233.59.12 <none> 8000/TCP 46h
kube-system kubernetes-dashboard ClusterIP 10.233.63.20 <none> 443/TCP 46h
But netchecker API gives the following answer :
[root#localhost ~]# curl http://X.X.X.X:31081/api/v1/connectivity_check
{"Message":"Connectivity check fails. Reason: there are absent or outdated pods; look up the payload","Absent":["netchecker-agent-hostnet-kk56x","netchecker-agent-hostnet-klldn","netchecker-agent-hostnet-r2vqs","netchecker-agent-hostnet-wqhjs"],"Outdated":["netchecker-agent-4jsgf","netchecker-agent-c9pcf","netchecker-agent-hostnet-jzbfv","netchecker-agent-vxgpf"]}
For an unknown reason, I cannot access the API from a cluster node with localhost, so I used a floating IP with OpenStack.
Here are some logs from the agent :
[centos#cl1-master-0 ~]$ sudo vi /var/log/pods/default_netchecker-agent-vjnwl_d8290268-3ea4-4e3c-acb4-295ab162a735/netchecker-agent/0.log
{"log":"I0701 13:04:01.814246 1 agent.go:135] Response status code: 200\n","stream":"stderr","time":"2020-07-01T13:04:01.81437579Z"}
{"log":"I0701 13:04:01.814272 1 agent.go:128] Sleep for 15 second(s)\n","stream":"stderr","time":"2020-07-01T13:04:01.814393199Z"}
{"log":"I0701 13:04:16.817398 1 agent.go:55] Send payload via URL: http://netchecker-service:8081/api/v1/agents/netchecker-agent-vjnwl\n","stream":"stderr","time":"2020-07-01T13:04:16.817786735Z"}
[centos#cl1-master-0 ~]$ sudo vi /var/log/pods/default_netchecker-agent-hostnet-klldn_d5fa6e72-885f-44e1-97a6-880a25e6d6d6/netchecker-agent/0.log
{"log":"E0701 13:05:22.804428 1 agent.go:133] Error while sending info. Details: Post http://netchecker-service:8081/api/v1/agents/netchecker-agent-hostnet-klldn: dial tcp 10.233.13.213:8081: i/o timeout\n","stream":"stderr","time":"2020-07-01T13:05:22.805138032Z"}
{"log":"I0701 13:05:22.804474 1 agent.go:128] Sleep for 15 second(s)\n","stream":"stderr","time":"2020-07-01T13:05:22.805190295Z"}
{"log":"I0701 13:05:37.807140 1 agent.go:55] Send payload via URL: http://netchecker-service:8081/api/v1/agents/netchecker-agent-hostnet-klldn\n","stream":"stderr","time":"2020-07-01T13:05:37.807309111Z"}
Logs from the server do not indicate any error.
I tried to check DNS resolve with the following :
[centos#cl1-master-0 ~]$ kubectl exec -it netchecker-agent-4jsgf -- /bin/sh
/ $ nslookup kubernetes.default
Server: 169.254.25.10
Address 1: 169.254.25.10
nslookup: can't resolve 'kubernetes.default'
[centos#cl1-master-0 ~]$ kubectl exec -it netchecker-agent-4jsgf -- cat /etc/resolv.conf
nameserver 169.254.25.10
search default.svc.cluster.local svc.cluster.local cluster.local openstacklocal
options ndots:5
169.254.25.10 is the IP of the nodelocaldns, but it doesn't seem to query the coredns service deployed.
When I use nslookup netchecker-service.default.svc.cluster.local 10.233.0.3, with the coredns IP, I get a correct answer.
What can be wrong with my configuration ?
Thanks in advance
UPDATE : The plugin Flannel has an issue and contains a fix to apply on all nodes of the cluster. Once done, the pods successfully report back to the netchecker server.
UPDATE : The plugin Flannel has an issue and contains a fix to apply on all nodes of the cluster. Once done, the pods successfully report back to the netchecker server.

Inquiring pod and service subnets from inside Kubernetes cluster

How can one inquire the Kubernetes pod and service subnets in use (e.g. 10.244.0.0/16 and 10.96.0.0/12 respectively) from inside a Kubernetes cluster in a portable and simple way?
For instance, kubectl get cm -n kube-system kubeadm-config -o yaml reports podSubnet and serviceSubnet. But this is not fully portable because a cluster may have been set up by another means than kubeadm.
kubectl get cm -n kube-system kube-proxy -o yaml reports clusterCIDR (i.e. pod subnet) and kubectl get pod -n kube-system kube-apiserver-master1 -o yaml reports the value
passed as command-line option --service-cluster-ip-range to kube-apiserver (i.e. service subnet). master1 stands for the name of any control plane node. But this seems a bit complex.
Is there a better way available e.g. with the Kubernetes 1.17 API?
I don't think it would be possible to obtain what you want in a portable and simple way.
If you don't specify Cidr's parameters it will assign default one.
As you have many ways to run kubernetes as unmanaged clusters like kubeadm, minikbue, k3s, micork8s or managed like Cloud providers (GKE, Azure, AWS) it's hard to find one way to list all cidrs in all environments. Another obstacle can be versions of Kubernetes or CNI.
In Kubernetes 1.17 Release notes you can find information that
Deprecate the default service IP CIDR. The previous default was 10.0.0.0/24 which will be removed in 6 months/2 releases. Cluster admins must specify their own desired value, by using --service-cluster-ip-range on kube-apiserver.
As example of Kubeadm: $ kubeadm init --pod-network-cidr 10.100.0.0/12 --service-cidr 10.99.0.0/12
There are a few ways to get this pod and service-cidr:
$ kubectl cluster-info dump | grep -E '(service-cluster-ip-range|cluster-cidr)'
"--service-cluster-ip-range=10.99.0.0/12",
"--cluster-cidr=10.100.0.0/12",
$ kubeadm config view | grep Subnet
podSubnet: 10.100.0.0/12
serviceSubnet: 10.99.0.0/12
But if you will check all pods in this cluster, some pods are starting with 192.168.190.X or 192.168.137.X
$ kubectl get pods -A -owide
NAMESPACE NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES
default nginx 1/1 Running 0 62m 192.168.190.129 kubeadm-worker <none> <none>
kube-system calico-kube-controllers-77c5fc8d7f-9n6m5 1/1 Running 0 118m 192.168.137.66 kubeadm-master <none> <none>
kube-system calico-node-2kx2v 1/1 Running 0 117m 10.128.0.4 kubeadm-worker <none> <none>
kube-system calico-node-8xqd9 1/1 Running 0 118m 10.128.0.3 kubeadm-master <none> <none>
kube-system coredns-66bff467f8-sgmkw 1/1 Running 0 120m 192.168.137.65 kubeadm-master <none> <none>
kube-system coredns-66bff467f8-t84ht 1/1 Running 0 120m 192.168.137.67 kubeadm-master <none> <none>
If you will describe any CNI pods you can find another CIDRs:
CALICO_IPV4POOL_CIDR: 192.168.0.0/16
For GKE example you will have:
node CIDRs
$ kubectl describe node | grep CIDRs
PodCIDRs: 10.52.1.0/24
PodCIDRs: 10.52.0.0/24
PodCIDRs: 10.52.2.0/24
$ gcloud container clusters describe cluster-2 --zone=europe-west2-b | grep Cidr
clusterIpv4Cidr: 10.52.0.0/14
clusterIpv4Cidr: 10.52.0.0/14
clusterIpv4CidrBlock: 10.52.0.0/14
servicesIpv4Cidr: 10.116.0.0/20
servicesIpv4CidrBlock: 10.116.0.0/20
podIpv4CidrSize: 24
servicesIpv4Cidr: 10.116.0.0/20
Honestly I don't think there is an easy and portable way to list all podCidrs and serviceCidrs in one simple command.

What happens when a service receives a request but has no ready pods?

Having a kubernetes service (of type ClusterIP) connected to a set of pods, but none of them are currently ready - what will happen to the request?
Will it:
fail eagerly
timeout
wait until a ready pod is available (or forever, whichever is earlier)
something else?
It will time out.
Kube-proxy pulls out the IP addresses from healthy pods and sets as endpoints of the service (backends). Also, note that all kube-proxy does is to re-write the iptables when you create, delete or modify a service.
So, when you send a request within your network and there is no one to reply, your request will timeout.
Deployed nginx service
[node1 ~]$ kubectl get svc
NAME TYPE CLUSTER-IP EXTERNAL-IP PORT(S) AGE
kubernetes ClusterIP 10.96.0.1 443/TCP 2h
my-nginx ClusterIP 10.100.1.134 80/TCP 9s
$ curl 10.100.1.134
curl: (7) Failed connect to 10.100.1.134:80; Connection refused
Deployed nginx deployment
$ kubectl create -f nginx-depl.yaml
$ kubectl get po
NAME READY STATUS RESTARTS AGE
my-nginx-f9945ffdd-2f77f 1/1 Running 0 1m
my-nginx-f9945ffdd-rk68v 1/1 Running 0 1m
$ curl 10.100.1.134
Welcome to nginx!
most likely you would get Connection refused error

kubernetes service IPs not reachable

So I've got a Kubernetes cluster up and running using the Kubernetes on CoreOS Manual Installation Guide.
$ kubectl get no
NAME STATUS AGE
coreos-master-1 Ready,SchedulingDisabled 1h
coreos-worker-1 Ready 54m
$ kubectl get cs
NAME STATUS MESSAGE ERROR
controller-manager Healthy ok
scheduler Healthy ok
etcd-0 Healthy {"health": "true"}
etcd-2 Healthy {"health": "true"}
etcd-1 Healthy {"health": "true"}
$ kubectl get pods --all-namespaces -o wide
NAMESPACE NAME READY STATUS RESTARTS AGE IP NODE
default curl-2421989462-h0dr7 1/1 Running 1 53m 10.2.26.4 coreos-worker-1
kube-system busybox 1/1 Running 0 55m 10.2.26.3 coreos-worker-1
kube-system kube-apiserver-coreos-master-1 1/1 Running 0 1h 192.168.0.200 coreos-master-1
kube-system kube-controller-manager-coreos-master-1 1/1 Running 0 1h 192.168.0.200 coreos-master-1
kube-system kube-proxy-coreos-master-1 1/1 Running 0 1h 192.168.0.200 coreos-master-1
kube-system kube-proxy-coreos-worker-1 1/1 Running 0 58m 192.168.0.204 coreos-worker-1
kube-system kube-scheduler-coreos-master-1 1/1 Running 0 1h 192.168.0.200 coreos-master-1
$ kubectl get svc --all-namespaces
NAMESPACE NAME CLUSTER-IP EXTERNAL-IP PORT(S) AGE
default kubernetes 10.3.0.1 <none> 443/TCP 1h
As with the guide, I've setup a service network 10.3.0.0/16 and a pod network 10.2.0.0/16. Pod network seems fine as busybox and curl containers get IPs. But the services network has problems. Originally, I've encountered this when deploying kube-dns: the service IP 10.3.0.1 couldn't be reached, so kube-dns couldn't start all containers and DNS was ultimately not working.
From within the curl pod, I can reproduce the issue:
[ root#curl-2421989462-h0dr7:/ ]$ curl https://10.3.0.1
curl: (7) Failed to connect to 10.3.0.1 port 443: No route to host
[ root#curl-2421989462-h0dr7:/ ]$ ip route
default via 10.2.26.1 dev eth0
10.2.0.0/16 via 10.2.26.1 dev eth0
10.2.26.0/24 dev eth0 src 10.2.26.4
It seems ok that there's only a default route in the container. As I understood it, the request (to default route) should be intercepted by the kube-proxy on the worker node, forwarded to the the proxy on the master node where the IP is translated via iptables to the masters public IP.
There seems to be a common problem with a bridge/netfilter sysctl setting, but that seems fine in my setup:
core#coreos-worker-1 ~ $ sysctl net.bridge.bridge-nf-call-iptables
net.bridge.bridge-nf-call-iptables = 1
I'm having a real hard time to troubleshoot, as I lack the understanding of what the service IP is used for, how the service network is supposed to work in terms of traffic flow and how to best debug this.
So here're the questions I have:
What is the 1st IP of the service network (10.3.0.1 in this case) used for?
Is above description of the traffic flow correct? If not, what steps does it take for a container to reach a service IP?
What are the best ways to debug each step in the traffic flow? (I can't get any idea what's wrong from the logs)
Thanks!
The Sevice network provides fixed IPs for Services. It is not a routeable network (so don't expect ip ro to show anything nor will ping work) but a collection iptables rules managed by kube-proxy on each node (see iptables -L; iptables -t nat -L on the nodes, not Pods). These virtual IPs (see the pics!) act as load balancing proxy for endpoints (kubectl get ep), which are usually ports of Pods (but not always) with a specific set of labels as defined in the Service.
The first IP on the Service network is for reaching the kube-apiserver itself. It's listening on port 443 (kubectl describe svc kubernetes).
Troubleshooting is different on each network/cluster setup. I would generally check:
Is kube-proxy running on each node? On some setups it's run via systemd and on others there is a DeamonSet that schedules a Pod on each node. On your setup it is deployed as static Pods created by the kubelets thrmselves from /etc/kubernetes/manifests/kube-proxy.yaml
Locate logs for kube-proxy and find clues (can you post some?)
Change kube-proxy into userspace mode. Again, the details depend on your setup. For you it's in the file I mentioned above. Append --proxy-mode=userspace as a parameter on each node
Is the overlay (pod) network functional?
If you leave comments I will get back to you..
I had this same problem, and the ultimate solution that worked for me was enabling IP forwarding on all nodes in the cluster, which I had neglected to do.
$ sudo sysctl net.ipv4.ip_forward=1
net.ipv4.ip_forward = 1
Service IPs and DNS started working immediately afterwards.
I had the same issue, turned out to be a configuration issue in kube-proxy.yaml For the "master" parameter I had the ip address as in - --master=192.168.3.240 but it actually required to be a url like - --master=https://192.168.3.240
FYI my kube-proxy sucessfully uses --proxy-mode=iptables (v1.6.x)

How to fix weave-net CrashLoopBackOff for the second node?

I have got 2 VMs nodes. Both see each other either by hostname (through /etc/hosts) or by ip address. One has been provisioned with kubeadm as a master. Another as a worker node. Following the instructions (http://kubernetes.io/docs/getting-started-guides/kubeadm/) I have added weave-net. The list of pods looks like the following:
vagrant#vm-master:~$ kubectl get pods --all-namespaces
NAMESPACE NAME READY STATUS RESTARTS AGE
kube-system etcd-vm-master 1/1 Running 0 3m
kube-system kube-apiserver-vm-master 1/1 Running 0 5m
kube-system kube-controller-manager-vm-master 1/1 Running 0 4m
kube-system kube-discovery-982812725-x2j8y 1/1 Running 0 4m
kube-system kube-dns-2247936740-5pu0l 3/3 Running 0 4m
kube-system kube-proxy-amd64-ail86 1/1 Running 0 4m
kube-system kube-proxy-amd64-oxxnc 1/1 Running 0 2m
kube-system kube-scheduler-vm-master 1/1 Running 0 4m
kube-system kubernetes-dashboard-1655269645-0swts 1/1 Running 0 4m
kube-system weave-net-7euqt 2/2 Running 0 4m
kube-system weave-net-baao6 1/2 CrashLoopBackOff 2 2m
CrashLoopBackOff appears for each worker node connected. I have spent several ours playing with network interfaces, but it seems the network is fine. I have found similar question, where the answer advised to look into the logs and no follow up. So, here are the logs:
vagrant#vm-master:~$ kubectl logs weave-net-baao6 -c weave --namespace=kube-system
2016-10-05 10:48:01.350290 I | error contacting APIServer: Get https://100.64.0.1:443/api/v1/nodes: dial tcp 100.64.0.1:443: getsockopt: connection refused; trying with blank env vars
2016-10-05 10:48:01.351122 I | error contacting APIServer: Get http://localhost:8080/api: dial tcp [::1]:8080: getsockopt: connection refused
Failed to get peers
What I am doing wrong? Where to go from there?
I ran in the same issue too. It seems weaver wants to connect to the Kubernetes Cluster IP address, which is virtual. Just run this to find the cluster ip:
kubectl get svc. It should give you something like this:
$ kubectl get svc
NAME CLUSTER-IP EXTERNAL-IP PORT(S) AGE
kubernetes 100.64.0.1 <none> 443/TCP 2d
Weaver picks up this IP and tries to connect to it, but worker nodes does not know anything about it. Simple route will solve this issue. On all your worker nodes, execute:
route add 100.64.0.1 gw <your real master IP>
this happens with a single node setup, too. I tried several things like reapplying the configuration and recreation, but the most stable way at the moment is to perform a full tear down (as described in docs) and put the cluster up again.
I use these scripts for relaunching the cluster:
down.sh
#!/bin/bash
systemctl stop kubelet;
docker rm -f -v $(docker ps -q);
find /var/lib/kubelet | xargs -n 1 findmnt -n -t tmpfs -o TARGET -T | uniq | xargs -r umount -v;
rm -r -f /etc/kubernetes /var/lib/kubelet /var/lib/etcd;
up.sh
#!/bin/bash
systemctl start kubelet
kubeadm init
# kubectl taint nodes --all dedicated- # single node!
kubectl create -f https://git.io/weave-kube
edit: I would also give other Pod networks a try, like Calico, if this is a weave related issue
The most common causes for this may be:
- presence of a firewall (e.g. firewalld on CentOS)
- network configuration (e.g. default NAT interface on VirtualBox)
Currently kubeadm is still alpha, and this is one of the issues that has already been reported by many of the alpha testers. We are looking into fixing this by documenting the most common problems, such documentation is going to be ready closer to beta version.
Right there exists a VirtualBox+Vargant+Ansible for Ubunutu and CentOS reference implementation that provides solutions for firewall, SELinux and VirtualBox NAT issues.
/usr/local/bin/weave reset
was the fix for me - Hope its useful - and yes make sure selinux is set to disabled
and firewalld is not running (on redhat / centos) releases
kube-system weave-net-2vlvj 2/2 Running 3 11d
kube-system weave-net-42k6p 1/2 Running 3 11d
kube-system weave-net-wvsk5 2/2 Running 3 11d