[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Flink 1.5 - Job fails to execute in multiple taskmanagers (parallelism > 1)

Thanks for reporting the issue Edward.

Taking a look at Netty SslHandler, it looks that we introduced this problem
with the update of the cipher algorithms [1]. Apparently, the SslHandler
wants to use inbound heap byte buffer when using a cipher suite with GCM
enabled [2]. This seems to be fixed with a later version of Netty 4 (we are
using 4.0.27 at the moment). The problem with the heap byte buffers are,
that our NettyBufferPool does not support the allocation of heap byte
buffers in order to have the memory consumption under control.

As a work around, you could set `security.ssl.algorithms` to
`TLS_RSA_WITH_AES_128_CBC_SHA` in the Flink configuration. That should make
it work again at the cost of using a cipher which is no longer recommended.

[1] https://issues.apache.org/jira/browse/FLINK-9310.


On Thu, May 24, 2018 at 8:49 PM, Edward Rojas <edward.rojascl@xxxxxxxxx>

> This may help to target the issue:
> If I let global ssl enabled, but I set taskmanager.data.ssl.enabled: false,
> it works again.
> --
> Sent from: http://apache-flink-mailing-list-archive.1008284.n3.nabble.com/