0

A h2o cluster runs on google cloud. Trying to connect to it from R Studio on Ubuntu using:

conn=h2o.connect(ip = "external ip", port=443, strict_version_check = FALSE, username = "username", password = "password", https = TRUE, insecure = TRUE)

throws the error: Can only start H2O launcher if IP address is localhost. Plus, the h2o command(s) also works extremely slow. We have tried reinstalling R so that some of the memory profile changes done earlier would not affect the h2o program.

bespectacled
  • 2,801
  • 4
  • 25
  • 35
  • 1
    what version of h2o are you using when you are connecting from your R client with h2o.connect()? Can you also specify if you are using the marketplace offering for google cloud or if it's an instance you stood up on your own? – Lauren Nov 02 '18 at 21:53
  • This question is lacking in details. What did you expect to happen? And how does that differ from what actually happened? – TomKraljevic Nov 03 '18 at 06:28
  • @Lauren we're using the marketplace offering in google cloud. The h2o version is 3.20.0.8 – bespectacled Nov 03 '18 at 12:05

1 Answers1

1

Unfortunately H2O-3 doesn't support using ip=external when you want to connect to h2o from R.

If you want to enable external access to an instance, you will need to upgrade to H2O-3 version 3.22.0.1 and use bind_to_localhost = false within h2o.init(), which will enable external access to the instance.

please see the R reference with details on bind_to_local: http://docs.h2o.ai/h2o/latest-stable/h2o-r/docs/reference/h2o.init.html

Lauren
  • 5,640
  • 1
  • 13
  • 19
  • Thanks. but i have connected to gcloud from R earlier. The problem seemed to be that, once the cluster is shutdown after work, since the external Ip is ephemeral, it did not exists anymore.Is there a way to create a H2O cluster and get a static ip to it? – bespectacled Nov 08 '18 at 06:35