Thanks, Robert! I tried the classloader.resolve.order: parent-first option but ran into SLF4J: Failed to load class "org.slf4j.impl.StaticLoggerBinder" errors (because I use logback so I followed https://nightlies.apache.org/flink/flink-docs-master/docs/deployment/advanced/logging/#configuring-logback and removed log4j-slf4j-impl from the classpath. But putting all my classes in lib/ instead of usrlib/ fixed that problem, and everything now runs fine. Thanks!
El vie, 28 ene 2022 a las 6:11, Robert Metzger (<metrob...@gmail.com>) escribió: > Hi Javier, > > I suspect that TwitterServer is using some classloading / dependency > injection / service loading "magic" that is causing this. > I would try to find out, either by attaching a remote debugger (should be > possible when executing in cluster mode locally) or by adding log > statements in the code, what the URI it's trying to load looks like. > > On the cluster, Flink is using separate classloaders for the base flink > system, and the user code (as opposed to executing in the IDE, where > everything is loaded from the same loader). Check out this page and try out > the config arguments: > https://nightlies.apache.org/flink/flink-docs-master/docs/ops/debugging/debugging_classloading/ > > > > On Wed, Jan 26, 2022 at 4:13 AM Javier Vegas <jve...@strava.com> wrote: > >> I am porting a Scala service to Flink in order to make it more scalable >> via running it in a cluster. All my Scala services extends a base Service >> class that extends TwitterServer ( >> https://github.com/twitter/twitter-server/blob/develop/server/src/main/scala/com/twitter/server/TwitterServer.scala) >> and that base class contains a lot of logic about resource initialization, >> logging, stats and error handling, monitoring, etc that I want to keep >> using in my class. I ported my logic to Flink sources and sinks, and >> everything worked fine when I ran my class in single JVM mode either from >> sbt or my IDE, Flink's jobmanager and taskmanagers start and run my app. >> But when I try to run my application in cluster mode, when launching my >> class with "./bin/standalone-job.sh start --job-classname" the >> jobmanager runs into a "IllegalArgumentException: URI is not hierarchical" >> error on initialization, apparently because TwitterServer is trying to load >> something from the class path (see attached full log). >> >> Is there anything I can do to run a class that extends TwitterServer in a >> Flink cluster? I have tried making my class not extend it and it worked >> fine, but I really want to keep using all the common infraestructure logic >> that I have in my base class that extends TwitterServer. >> >> Thanks! >> >
