[
https://issues.apache.org/jira/browse/FLINK-23945?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17404330#comment-17404330
]
Chesnay Schepler commented on FLINK-23945:
------------------------------------------
I suppose it's expected that is interprets the endpoint as the bucket because
s3.endpoint is still set and evaluated by the filesystem, so in the end you
have the endpoint twice in the URL.
It could make sense that s3:///flink/recovery, fails since s3.endpoint contains
a trailing slash and is injected later on as-is (I suppose?), you could
theoretically end up with an incorrect number of slashes.
I couldn't find this exact issue in the web search; I found _similar_ issues
because of wrongly formatted/encoded credentials however.
This issue is a bit of a mystery TBH because we have e2e tests that do exactly
the same thing.
> can not open dashboard
> -----------------------
>
> Key: FLINK-23945
> URL: https://issues.apache.org/jira/browse/FLINK-23945
> Project: Flink
> Issue Type: Bug
> Components: Connectors / FileSystem, Deployment / Kubernetes
> Affects Versions: 1.12.3
> Reporter: 钟洋洋
> Priority: Minor
>
> Cannot deploy normally using k8s native session mode, but no error log is
> found
>
> Details are in the comments section
>
>
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
