Marek Novotný

It fails on creation of MOJO model in H2O-3, when serializing it's meta-data. I can not understand why the core meta data of the model became so big. Would it...

Hi @cliu-sift, I'm struggling to replicate the problem on the dataset that you shared. Here is the script that I'm running: ``` from pysparkling.ml import H2OGLM from pyspark.ml import Pipeline...

Hi @cliu-sift, any hints what could be doing wrong in reproducing the problem?

@cliu-sift I've reproduced the problem and created the [PUBDEV-8838](https://h2oai.atlassian.net/browse/PUBDEV-8838) ticket for H2O-3 team.

Hi @gurumoorthy208524 , From the logs you provided, you try to connect to k8s service `h2o-service.sparkling-water.svc.cluster.local`. Isn't your service named differently? (`data-py-9a30b280e693f596-driver-svc.test.svc`) and thus set SW property `spark.ext.h2o.cloud.representative` to `data-py-9a30b280e693f596-driver-svc.test.svc:54321`?

> Note: we are using preemptive nodes for both spark & h2o. @gurumoorthy208524 This a big source of problems. H2O-3 is not fault-tolerant by design due to performance reasons. (all...

@kaleex I looked into this problem some time ago, but I couldn't reproduce it. Can you share the version of your Databricks runtime and the spark properties that you have...

@gurumoorthy208524, it's difficult to help you if we don't have a full context. For internal backend, Please always send us the full logs from driver and executors, not only the...

@gurumoorthy208524 Several observations from the logs: - The frame doesn't seem to be parsed correctly, min and max values from the frame statistics: ``` 2022-05-30T11:15:54.534422466Z 05-30 11:15:54.533 10.171.128.84:54321 14 1868849-69...

Hi @bradmiro, If you want to build Sparkling Water on your own, I recommend to set git to point to a release tag: ```git checkout RELEASE-3.30.0.6-1``` If you use `rel-*`...