:bug: Stuck at "Waiting for Kagenti ADK platform to be ready..."

[esnible]

Describe the bug

I attempted to install Kagenti ADK using sh -c "$(curl -LsSf https://raw.githubusercontent.com/kagenti/adk/main/install.sh)"

After a few minutes I see

Installing kagenti platform (operator + backend) [DONE] 
Waiting for DNS to be ready [DONE] 
Forwarding VM services to host [DONE] 
⠹ Waiting for Kagenti ADK platform to be ready...

This doesn't finish. Eventually, after an over-all timeout, I see "❗ WARNING: Platform start failed. You can retry with kagenti-adk platform start."

Steps To Reproduce

I suspect the problem is that I already have Kagenti running on port 8080 because I installed it with deployments/ansible/run-install.sh

1. Clone https://github.com/kagenti/kagenti
2. deployments/ansible/run-install.sh
3. sh -c "$(curl -LsSf https://raw.githubusercontent.com/kagenti/adk/main/install.sh)"

Expected Behavior

It would be helpful for the end-user to receive more guidance.

• "Waiting for Kagenti ADK platform to be ready..." should eventually timeout
• "Waiting for Kagenti ADK platform to be ready..." should tell the user what it is trying to do
• Perhaps it should refuse to install if there is already processes listening on 8080
• ADK is hard-coded to 8080, and Kagenti's own installer is also hard-coded to that.
  • e.g.

    adk/apps/adk-cli/src/kagenti_cli/commands/platform.py

    Line 923 in 9fe61e2

    (await client.get("http://adk-api.localtest.me:8080/healthcheck")).raise_for_status()

  • The MCP Gateway's installation scripts allow the port to be customized by using export KIND_HOST_PORT_MCP_GATEWAY=8090. Perhaps a similar config option could be used with ADK to allow a user to install both environments together.

Additional Context

No response

[esnible]

Maybe I am mistaken about the reason. I just tried installing on a second Mac, one that does not have Kagenti or Kind installed.

After 20 minutes the install fails, with the same message. The problem might be related to DNS:

curl -v http://adk-api.localtest.me:8080/healthcheck
* Could not resolve host: adk-api.localtest.me
* Closing connection
curl: (6) Could not resolve host: adk-api.localtest.me
edsnible@MacBookPro /tmp % nslookup adk-api.localtest.me
Server:		2600:4040:96b3:1a00::1
Address:	2600:4040:96b3:1a00::1#53

Non-authoritative answer:
*** Can't find adk-api.localtest.me: No answer

[JanPokorny]

@esnible *.localtest.me should resolve to 127.0.0.1

Are you behind some corporate VPN that could tamper with DNS?

The "correct" way would be to use adk-api.localhost for the domain, but kagenti already uses *.localtest.me which is a third-party domain resolving to 127.0.0.1 so we used it for ADK too.

[esnible]

@JanPokorny I have tried this both at home at at my IBM location.

At my work location, nslookup adk-api.localtest.me does give 127.0.0.1. The problem is not the IPv6 DNS.