PubSub Lite DataFlow Errors - ManagedChannelImpl was not shutdown properly

[GLStephen]

When running PubSubLite as a subscriber in a dataflow job I'm receiving the following error repeatedly.

"*~*~*~ Channel ManagedChannelImpl{logId=845, target=us-east1-pubsublite.googleapis.com:443} was not shutdown properly!!! ~*~*~*
    Make sure to call shutdown()/shutdownNow() and wait until awaitTermination() returns true."

The connection is handled in the Pipeline. Do I need to get a reference to the connection and shut it down somewhere? This feels like it should be managed somewhere lower level in this deployment context.

Environment details

1. OS type and version: Windows -> Dataflow GCP
2. Java version: 1.8
3. Beam 1.28
4. pubsublite version(s):

<dependency>
      <groupId>com.google.cloud</groupId>
      <artifactId>pubsublite-beam-io</artifactId>
      <version>0.11.1</version>
    </dependency>

Steps to reproduce

1. Create PubSubLit Subscriber in DataFlow Job

[GLStephen]

Possibly related. It seems that a Job using this library will never fully drain. It has to be canceled. That's a bit of a major roadblock for production deployments.

[dpcollins-google]

I'm receiving the following error repeatedly.

Yeah this is a known issue. I think I know which channel this is; the ones from ServiceClients. Its going to be a bit difficult to fix, but shouldn't cause any issues other than noise

a Job using this library will never fully drain

This is more concerning. I'm going to reach out to the dataflow team and investigate this.

[dpcollins-google]

From Dataflow team, I've added the following text to the README. Unfortunately it looks like the v1 runner does not support drains with SplittableDoFns:

Only Dataflow runner v2 supports drain on PubsubLiteIO. Please contact dataflow customer support to request adding projects into the allowlist and launch the pipeline with --experiments=use_runner_v2. Using runner v2 requires 2.29.0 or above.

[dpcollins-google]

Follow up from dataflow team to edit to the following:

NOTE: Drain is not currently supported on PubsubLiteIO. Once Beam 2.29.0 is released, this will be supported
only when --experiments=use_runner_v2 is enabled. Please contact dataflow customer support to request adding projects
into the allowlist and launch the pipeline with --experiments=use_runner_v2.

[GLStephen]