Forgot to mention the Flink version is 1.13.2 and we use kubernetes native
mode
On Thu, May 12, 2022 at 9:18 PM tao xiao <xiaotao...@gmail.com> wrote:
> Hi team,
>
> I met a weird issue when a job tries to recover from JM failure. The
> success checkpoint before JM crashed is 41205
>
> ```
>
> {"log":"2022-05-10 14:55:40,663 INFO
> org.apache.flink.runtime.checkpoint.CheckpointCoordinator [] - Completed
> checkpoint 41205 for job 00000000000000000000000000000000 (9453840 bytes in
> 1922 ms).\n","stream":"stdout","time":"2022-05-10T14:55:40.663286893Z"}
>
> ```
>
> However JM tries to recover the job with an old checkpoint 41051 which
> doesn't exist that leads to unrecoverable state
>
> ```
>
> "2022-05-10 14:59:38,949 INFO
> org.apache.flink.runtime.checkpoint.DefaultCompletedCheckpointStore [] -
> Trying to retrieve checkpoint 41051.\n"
>
> ```
>
> Full log attached
>
> --
> Regards,
> Tao
>
--
Regards,
Tao
