h2o.set_s3_credentials should support endpoint and region

[hutch3232]

Is your feature request related to a problem? Please describe.
It's convenient that h2o supports changing keys within a session using h2o.set_s3_credentials, but the endpoint and region values are set before launching h2o. It would be nice if those were easily configurable within a session.

Describe the solution you'd like
Two new arguments, perhaps endpoint and region, to h2o.set_s3_credentials. Probably defaulting to NULL or something to not break existing code.

Describe alternatives you've considered
Sparkling water supports this because you can configure bucket-level access which is super nice. That kind of an approach would be even better, if possible.

Additional context
This PR exposed the endpoint/region args: #164 to java

[hutch3232]

It would also be a small but nice enhancement to replace print with message. That would be a more typical R approach and supports suppressing it.

h2o-3/h2o-r/h2o-package/R/import.R

Line 456 in 733c801

print("Credentials successfully set.")

[krasinski]

@hutch3232 thank you for reporting that, that is a very valid point
before that gets implemented you can try playing with "sys.ai.h2o.persist.s3.endPoint" and "sys.ai.h2o.persist.s3.region" properties

probably something like options(java.parameters = "-Dsys.ai.h2o.persist.s3.endPoint=https://url:9000")

[hutch3232]

It doesn't seem to be getting picked up. My test:

h2o.init(...)
options(java.parameters = "-Dsys.ai.h2o.persist.s3.endPoint=...")
h2o.set_s3_credentials(...)
df <- h2o.importFolder(...) # error with connection reset

# restart R session

h2o.init(..., jvm_custom_args = "-Dsys.ai.h2o.persist.s3.endPoint=...")
h2o.set_s3_credentials(...)
df <- h2o.importFolder(...) # works!

[krasinski]

so h2o.init method works? that's great to hear

[hutch3232]

Yep! I still think it would be nice if it could be switched/set mid-session via h2o.set_s3_credentials.