You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I've a Spark application with dependencies on H2O Sparkling Water jars
// H2O Sparkling Water version (spark 2.4)
val h2oVersion = "3.32.0.1-2-2.4"
lazy val h2oLibs = Seq(
"org.apache.spark" %% "spark-repl" % sparkVersion,
"ai.h2o" % "sparkling-water-package_2.11" % h2oVersion //exclude("ai.h2o", "sparkling-water-api-generation")
)
When I try to start it on k8s, the Spark executor starts properly (see logs from executor container below) and I can open a port-forward connection to it and access flow UI.
But the driver node gets stack in infinity loop and I see in the logs the following
2020-12-03T17:04:59.532+0000 level=INFO thread=main logger=ai.h2o.sparkling.backend.utils.RestApiUtils
H2O node http://10.1.0.22:54321/3/verifyWebOpen successfully responded for the GET.
2020-12-03T17:04:59.553+0000 level=INFO thread=main logger=ai.h2o.sparkling.backend.utils.RestApiUtils
H2O node http://10.1.0.22:54321/3/verifyVersion?referenced_version=3.32.0.1 successfully responded for the GET.
2020-12-03T17:04:59.636+0000 level=INFO thread=main logger=ai.h2o.sparkling.backend.utils.RestApiUtils
H2O node http://10.1.0.22:54321/3/Cloud successfully responded for the GET.
2020-12-03T17:04:59.713+0000 level=INFO thread=main logger=ai.h2o.sparkling.backend.utils.RestApiUtils
H2O node http://10.1.0.22:54321/3/LogLevel successfully responded for the GET.
2020-12-03T17:04:59.951+0000 level=INFO thread=main logger=ai.h2o.sparkling.backend.utils.RestApiUtils
H2O node http://10.1.0.22:54321/99/Rapids successfully responded for the POST.
2020-12-03T17:05:00.011+0000 level=WARN thread=main logger=ai.h2o.sparkling.backend.utils.ProxyStarter
Tried using port 54321 for Flow proxy, but port was already occupied!
Apparently, when SW tries to check for available ports it uses the Spark driver host but later jetty server binds to 0.0.0.0
I've a Spark application with dependencies on H2O Sparkling Water jars
When I try to start it on k8s, the Spark executor starts properly (see logs from executor container below) and I can open a port-forward connection to it and access flow UI.
But the driver node gets stack in infinity loop and I see in the logs the following
Apparently, when SW tries to check for available ports it uses the Spark driver host but later jetty server binds to 0.0.0.0
See here for detailed discussion #2414
The text was updated successfully, but these errors were encountered: