You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@avro.apache.org by "Ryan Skraba (Jira)" <ji...@apache.org> on 2021/06/30 13:45:00 UTC

[jira] [Updated] (AVRO-3165) [avro-ipc-netty] NioEventLoopGroup leak in NettyTransceiver

     [ https://issues.apache.org/jira/browse/AVRO-3165?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Ryan Skraba updated AVRO-3165:
------------------------------
    Fix Version/s: 1.11.0
           Status: Patch Available  (was: Open)

> [avro-ipc-netty] NioEventLoopGroup leak in NettyTransceiver
> -----------------------------------------------------------
>
>                 Key: AVRO-3165
>                 URL: https://issues.apache.org/jira/browse/AVRO-3165
>             Project: Apache Avro
>          Issue Type: Bug
>          Components: java
>    Affects Versions: 1.10.2, 1.10.1, 1.10.0
>            Reporter: Benjamin Leber
>            Assignee: Benjamin Leber
>            Priority: Major
>              Labels: ipc, netty
>             Fix For: 1.11.0
>
>         Attachments: image-2021-06-30-07-53-20-767.png
>
>
> The NettyTransceiver is used on server and client side. While it requires a worker group on client side this is not necessary on server side.
> But as the worker group is initialized during the variable declaration. It is also created on the server for every new connection, but it gets never closed. This causes high off heap memory and always increasing open file descriptors.
> !image-2021-06-30-07-53-20-767.png!
> Moving the creation of the workger group into the constructor which is called on client side, solves the problem (deployed after the second 12:00 in above graph)
> I will provide a pr to fix.



--
This message was sent by Atlassian Jira
(v8.3.4#803005)