[ 
https://issues.apache.org/jira/browse/FLINK-15871?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17060620#comment-17060620
 ] 

Canbin Zheng commented on FLINK-15871:
--------------------------------------

Hi, [~fly_in_gis]!

The init container is a basic design since we need it to download dependencies 
in the per job or the so-called application mode, and the users do not have to 
do much on using the init container. So it's reasonable that we introduce this 
feature for the codebase.

As to the sidecar container, I think it is an advanced feature instead of a 
basic one. We'd better not introduce explicit support for it and leave the 
users to enable such a feature via the pod template.

For the scenarios you mentioned for the sidecar container, I would like to 
share some information:

> Collect logs

Usually, we do not use sidecar containers to collect logs to avoid efficiency 
issues. The more general way is to start a Service on the host to collect all 
the logs for all the Pods located on the same host. In fact, the latter 
solution improves efficiency since we do not start a sidecar process for every 
pod.

 

> With debugging image which contains lots of tools and help to debug the 
> network problems

We could add tools into the basic image used by the main Container of the JM 
and TMs.

 

 

> Support to start sidecar container
> ----------------------------------
>
>                 Key: FLINK-15871
>                 URL: https://issues.apache.org/jira/browse/FLINK-15871
>             Project: Flink
>          Issue Type: Sub-task
>          Components: Deployment / Kubernetes
>    Affects Versions: 1.10.0
>            Reporter: Yang Wang
>            Priority: Major
>             Fix For: 1.11.0
>
>
> >> How does sidecar container work?
> A sidecar container is running beside the Jobmanager and TaskManager 
> container. It could be used to collect log or debug some problems. For 
> example, when we configure the sidecar container to fluentd and share the 
> TaskManager log with volume, then it could be used to upload the logs to 
> HDFS, elastic search, etc. Also we could start a sidecar container with 
> debugging image which contains lots of tools and help to debug the network 
> problems.



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

Reply via email to