Thanks for the response Guowei.
* Tried a telnet to the jobmanager host:port and I get “127.0.0.1:8086: nodename nor servname provided, or not known” which suggests that the network access is fine ? * I resubmitted the word count example and it ran fine to completion. For the pulsar script, I have also tried localhost, and the local LAN Ips as jobmanager host configuration in conf/flink.yaml and all of them end with the same result. I have also tried this with Pulsar 2.8.0 and it did have issues with “shared” subscription type (Get a “transactions not enabled” error in spite of enabling transactions in 2.8.0 broker). When I change the subscription type to “Exclusive” it exhibits the same behavior as the Pulsar 2.9.1 version. i.e. The job manager submission fails. (in both 2.8.0 pulsar and 2.9.1 pulsar) Regards, Ananth From: Guowei Ma <guowei....@gmail.com> Date: Monday, 21 February 2022 at 4:57 pm To: Ananth Gundabattula <agundabatt...@darwinium.com> Cc: user@flink.apache.org <user@flink.apache.org> Subject: Re: Pulsar connector 2.9.1 failing job submission in standalone. Hi, Ansanth I don't see any error logs on the server side, so it's hard to tell what the specific problem is. From the current log, there are two things to try first: 1. From the client's log, it is a 5-minute timeout, so you can telnet 127.0.0.1:8086<http://127.0.0.1:8086> to see if there is a problem with the local network 2. From the log on the server side, there is no job submission at all. You can try to submit the wordcount example again when submitting the pulsar example fails. So as to rule out whether the session cluster is inherently problematic. Best, Guowei On Mon, Feb 21, 2022 at 9:48 AM Ananth Gundabattula <agundabatt...@darwinium.com<mailto:agundabatt...@darwinium.com>> wrote: Hello All, I have a Pyflink script that needs to read from Pulsar and process the data. I have done the following to implement a prototype. 1. Since I need Pyflink way to connect to Pulsar , I checked out the code from master branch as advised in a different thread. (PyFlink Pulsar connector seems to be slated for 1.15 release) 2. I built the Flink source. 3. I am using the following location as FLINK_HOME under the source: flink-dist/target/flink-1.15-SNAPSHOT-bin/flink-1.15-SNAPSHOT 4. The python pyflink wheels have been appropriately installed in the right python conda environment. 5. I copied the flink-sql-connector-pulsar-1.15-SNAPSHOT.jar into the $FLINK_HOME/lib folder. 6. I started the standalone cluster by running bin/start-cluster.sh 7. I submit my test script by using bin/flink run –python … 8. If am launching the the word_count example in flink documentation, everything runs fine and it completes successfully. 9. However, if the script involves the Pulsar connector, the logs show that the Flink client codebase is not able to submit the job to the Jobamanger. 10. It ultimately dies with a Channel Idle exception. (See this in DEBUG mode of the logs). I am attaching the logs for reference. I am trying this on OSx. Please note that the classic word_count script works fine without any issues and I see the job submission failures on the client only when the pulsar source connector is in the script. I have also added the logs for the standalone session job manager.I am also attaching the script for reference. Could you please advise what can I do to resolve the issue. (Will raise an JIRA-Issue if someone thinks it is a bug). Regards, Ananth
