Hi All, I am exploring `Disaggregated State Manage...
# troubleshooting
s
Hi All, I am exploring
Disaggregated State Management
with
ForSt
state backend on
Flink 2.0
, even with the following config, during Flink app restoring from state, I see the the full checkpoints are downloaded into the taskmanager pod, do you know what I might be missing?
Copy code
state.backend.type: "forst"
    table.exec.async-state.enabled: "true"
    execution.checkpointing.incremental: "true"
    table.exec.mini-batch.enabled: "false"
    table.optimizer.agg-phase-strategy: "ONE_PHASE"
p
are you also setting
execution.checkpointing.dir
?
✅ 1
Perhaps
state.backend.forst.sync.enforce-local: false
may also help
✅ 1
s
I tried setting both of these, however, the outcome was the same, I am wondering if it has to do with using
inner join
Flink SQL, the documentation although, says
joins
are in general support async state, the checkpoint files in s3 has names like
op_AsyncStateStreamingJoinOperator_*
, which might indicate that it is infact using
async
state, however I am confused by the behavior