No route to host/connection refused about frakti HOT 12 CLOSED

@DarkBlaez How the dashboard is deployed, It works ok in my settings:

kubectl create -f https://raw.githubusercontent.com/kubernetes/dashboard/master/src/deploy/kubernetes-dashboard.yaml

Wait a while for pod up and then visit http://hostip:service-nodePort.

from frakti.

My question in reply is, are you starting from scratch, clean Centos 7.3, and following the deploy or allinone.sh, then adding a second node? Each time I try this and the reason I've found the issues is from doing a clean install. I do this as common practice. Otherwise if I develop code and check it in, there is no guarantee a fresh install will work since I can't account for prior dependencies being met do to past installs.

from frakti.

Hmm, right. We are only support bridge plugin today, so it is expected for users to configure subnet routes on each host. I will add the guide later.

Could you add subnet routes on each host and try again?

from frakti.

I would need install guide to step me through adding subnet route to each host. Have not done it prior

from frakti.

@DarkBlaez Filed PR #137 for setting subnet routes, could you have a try on it?

from frakti.

I will start with fresh VMs and let you know results. By the way in the routes where are 10.140.0.2 and 10.140.0.3 coming from?

from frakti.

Hmm, sorry for the mistake. They are IPs of each host. Made a fix in the PR.

from frakti.

Just for continuity in documentation, deploy.md will need the updated steps which were included in the allinone.sh

Centos
-baseurl=http://yum.kubernetes.io/repos/kubernetes-el7-x86_64-unstable
+baseurl=http://yum.kubernetes.io/repos/kubernetes-el7-x86_64

And the appropriate changes for Ubuntu just to maintain continuity in documentation

Also I assume that the CNI entries for master and each subsequent node should be different?

So on master:
"ipam": {
"type": "host-local",
"subnet": "10.244.1.0/24",
"routes": [
{ "dst": "0.0.0.0/0" }

Then node 1:
"ipam": {
"type": "host-local",
"subnet": "10.244.2.0/24",
"routes": [
{ "dst": "0.0.0.0/0" }
and so forth for each additional node, would this be a correct assumption?

Thanks
DB

from frakti.

@DarkBlaez Thanks. Would you like to file a PR for this?

from frakti.

Within the allinone.sh script I have noticed that when it runs for some reason the "setenforce 0" does not actually disable selinux. This may be do to the way the script runs. If I do each step manually then this gets set properly. I am going to step through this again doing it manually and accounting for the changes which were applied to the allinone.sh script which differs from deploy.md. I recreate fresh VMs and let you know the result

DB

from frakti.

Another item to note, this change is in the deploy.md but not in the allinone.sh. Is this needed for Centos 7?

deploy.md:
sed -i 's/native.cgroupdriver=systemd/native.cgroupdriver=cgroupfs/g' /usr/lib/systemd/system/docker.service

allinone.sh:

May be helpful to disable selinux. Using setenforce 0 does not seem to stick on reboot and upon reboot selinux is enforcing. I think some guides indicated either disabling selinux through /etc/selinux/config or completely disabling the service. I will further test and keep you posted before further changes are made

DB

from frakti.

Whether I install manually or through the allinone.sh the result is the same. Likewise, the ip routes are simplistic and would work great on a set of VMs sitting on the same subnet. A situation which is ideal at best. I will put this on hold and come back to the project once things are further along. I think once this supports plugable networks like Calico, Flannel , etc I think things will go smoother

from frakti.