Hi Jared,
I just wanted to follow up on this problem that you reported. Are there any new
insights about this problem from your debugging efforts and does it still
exists for you?
Best,
Stefan
> Am 09.07.2017 um 18:37 schrieb Jared Stehler
> :
>
> We are using the rocksDB state backend. We h
We are using the rocksDB state backend. We had not activated incremental
checkpointing, but in the course of debugging this, we ended up doing so, and
also moving back to S3 from EFS as it appeared that EFS was introducing large
latencies. I will attempt to provide some profiler data as we are a
On Mon, Jul 3, 2017 at 12:02 PM, Stefan Richter
wrote:
> Another thing that could be really helpful, if possible, can you attach a
> profiler/sampling to your job manager and figure out the hotspot methods
> where most time is spend? This would be very helpful as a starting point
> where the probl
Hi,
I have two quick questions about this problem report:
1) Which state backend are you using?
2) In case you are using RocksDB, did you also activate incremental
checkpointing when moving to Flink 1.3.
Another thing that could be really helpful, if possible, can you attach a
profiler/samplin
We’re seeing our external checkpoints directory grow in an unbounded fashion…
after upgrading to Flink 1.3. We are using Flink-Mesos.
In 1.2 (HA standalone mode), we saw (correctly) that only the latest external
checkpoint was being retained (i.e., respecting state.checkpoints.num-retained
def
