[ https://issues.apache.org/jira/browse/BEAM-8112?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Ismaël Mejía updated BEAM-8112: ------------------------------- Status: Open (was: Triage Needed) > Support passing stateBackend through pipeline options in python sdks > -------------------------------------------------------------------- > > Key: BEAM-8112 > URL: https://issues.apache.org/jira/browse/BEAM-8112 > Project: Beam > Issue Type: New Feature > Components: runner-flink > Reporter: Catlyn Kong > Priority: Major > > Currently the only way for python sdks to instruct flink to use a > StateBackend different than the default (MemoryStateBackend) would be to > specify state.backend in flink-conf.yaml, which creates the limitation of > using the same statebackend for every job running on the same flink cluster. > Ideally we should be able to pass it in to flink runner through > PipelineOptions. Here's the error it spits out when I flag > --state_backend=RocksDBStateBackend: > > {code:java} > RuntimeError: Pipeline failed in state FAILED: > com.fasterxml.jackson.databind.exc.InvalidDefinitionException: Cannot > construct instance of `org.apache.flink.runtime.state.StateBackend` (no > Creators, like default construct, exist): abstract types either need to be > mapped to concrete types, have custom deserializer, or contain additional > type information > at [Source: (String)""RocksDBStateBackend""; line: 1, column: 1] > {code} > Acceptance Criteria: > Flink StateBackend is configurable via command line options from python. > -- This message was sent by Atlassian Jira (v8.3.2#803003)