Hi Vasily,
I haven't tested the stare recovery under YARN setup. But in case of
stand-alone Flink cluster setup, I needed to run the application with proper
open-checkpoint recovery directory (whose name stars with 'chk-') passed as -s
parameter value. This was the only way I could recover my a
Hi, Maxim.
My console-consumer command:
kafka-console-consumer --zookeeper ... --topic test --from-beginning
--isolation-level read_committed
It works perfectly well with manually written kafka producer - it reads
data only after commitTransaction.
On Fri, 2 Aug 2019 at 14:19, Maxim Parkachov
Hi Vasily,
as far as I know, by default console-consumer reads uncommited.
Try setting isolation.level to read_committed in console-consumer
properties.
Hope this helps,
Maxim.
On Fri, Aug 2, 2019 at 12:42 PM Vasily Melnik <
vasily.mel...@glowbyteconsulting.com> wrote:
> Hi, Eduardo.
> Maybe i
Hi, Eduardo.
Maybe i should describe experiment design precisely :
1) I run Flink on YARN (YARN Session method).
2) I do not stop/cancell application, i just kill TaskManager process
3) After that YARN creates another TaskManager Process and auto checkpoint
restore from HDFS happens.
That's why i
Hi Vasily,
You're probably executing this from your IDE or from a local Flink cluster
without starting your job from a checkpoint.
When you start your Flink job for the second time you need to specify the
path to the latest checkpoint as an argument, otherwise Flink will start
from scratch.
You'
I,m trying to test Flink 1.8.0 exactly-once semantics with Kafka Source and
Sink:
1. Run flink app, simply transferring messages from one topic to another
with parallelism=1, checkpoint interval 20 seconds
2. Generate messages with incrementing integer numbers using Python
script each
