[FLINK-10445] CLI stuck with (potentially incorrect) SSL config

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Won't Fix
• Affects Version/s: 1.5.4
• Component/s: Command Line Client, Runtime / REST
• Assignee: Unassigned
• Reporter: Nico Kruber
• Created: 26/Sep/18 16:32
• Updated: 15/Jan/21 11:50
• Resolved: 15/Jan/21 11:50

Description

I created a keystore with

keytool -genkeypair -alias flink.internal -keystore internal.keystore -dname "CN=flink.internal" -storepass internal_store_password -keypass internal_key_password -keyalg RSA -keysize 4096

and adapted flink-conf.yaml to point to this with the following configuration:

security.ssl.enabled: true
web.ssl.enabled: false
security.ssl.keystore: /path/to/internal.keystore
security.ssl.keystore-password: internal_store_password
security.ssl.key-password: internal_key_password
security.ssl.truststore: /path/to/internal.keystore
security.ssl.truststore-password: internal_store_password

Now I know that this setup is supposed to work with Flink 1.6+ but probably wrong for 1.5. However, I was able to start up a Flink cluster and tried to submit a job via

./bin/flink run -p 1 ./examples/streaming/WordCount.jar --input /home/nico/Projects/flink/LICENSE

and nothing happened. The CLI was stuck at

Starting execution of program
Printing result to stdout. Use --output to specify output path.

and the job did not show up in the Web UI. I can also not find anything related to the job submission in the JM log.

Debugging into CliFrontend, it eventually enters RestClient#submitRequest but doesn't seem to get a response. RestClient.MultipartRequest#writeTo does add a listener on the future and debugging in there, the ignored object handed to the lambda is a DefaultChannelPromise@1a884b08(failure(java.nio.channels.ClosedChannelException) which looks like we should react on. However, that alone is not really enough information on the cause of the error and requires further investigation.