cli failed to connect

[lluunn]

I was trying the getting start guide.
I can see the pod using kubectl.

But

katib-cli -s gke-test-katib-default-pool-b88188d2-jnvp:30678 Getstudies
2018/05/09 11:38:13 connecting gke-test-katib-default-pool-b88188d2-jnvp:30678
2018/05/09 11:38:14 grpc: addrConn.resetTransport failed to create client transport: connection error: desc = "transport: dial tcp: lookup gke-test-katib-default-pool-b88188d2-jnvp on 127.0.0.1:53: no such host"; Reconnecting to {gke-test-katib-default-pool-b88188d2-jnvp:30678 <nil>}
2018/05/09 11:38:14 GetStudy failed: rpc error: code = 14 desc = grpc: the connection is unavailable

11:38:14lunkai@None:katib$ katib-cli
2018/05/09 11:38:34 connecting 127.0.0.1:6789
2018/05/09 11:38:34 Method not found: 

Do we need to configure cli somehow?

[gaocegege]

Do you download the CLI from GitHub or build it on your own?

[lluunn]

download from github, as https://github.com/kubeflow/katib/blob/master/docs/getting-start.md#install-the-system-and-cli

[gaocegege]

Then you need to checkout v0.1.0-alpha to have a try. We do not have a CD process for CLI so it is version specific.

[gaocegege]

Wait a minute, @YujiOshima Does our images have the tag v0.1.0-alpha or v0.1.1-alpha for test?

[YujiOshima]

@lluunn @gaocegege If you are using latest tagged katib images, you need to use v0.1.1-alpha.

[lluunn]

I am using v0.1.1-alpha now. Got a different error

13:28:10lunkai@None:katib$ katib-cli -s gke-test-katib-default-pool-b88188d2-jnvp:30678 get studies
2018/05/22 13:28:41 GetStudy failed: rpc error: code = Unavailable desc = all SubConns are in TransientFailure, latest connection error: 
connection error: desc = "transport: Error while dialing dial tcp: lookup gke-test-katib-default-pool-b88188d2-jnvp on 127.0.0.1:53: no such host"

It seems it's trying to find the pool on 127.0.0.1?

[YujiOshima]

It looks problem of DNS setting on your host.
Can you nslookup gke-test-katib-default-pool-b88188d2-jnvp from your host?

[lluunn]

nslookup gke-test-katib-default-pool-b88188d2-jnvp
Server:         127.0.0.1
Address:        127.0.0.1#53

** server can't find gke-test-katib-default-pool-b88188d2-jnvp: NXDOMAIN

To clarify, I am deploying katib to a GKE cluster, and calling katib-cli from my laptop.
Is it expected to work this way?
Do a need to setup anything for my cluster?

[YujiOshima]

You failed to resolve the name gke-test-katib-default-pool-b88188d2-jnvp.
Try to use the IP address of the node.
And in GKE, you should add a firewall rule to accessing to port 30678.
e.g. gcloud compute firewall-rules create katibservice --allow tcp:30678

[lluunn]

Working now, thank you!
I will sent out a PR to fix getting-started.md