[BEAM-10113] PubSubIO readMessagesWithMessageId() breaks the payload encoding when using DataflowRunner

• Type: Bug
• Status: Open
• Priority: P3
• Resolution: Unresolved
• Affects Version/s: 2.20.0
• Fix Version/s: Not applicable
• Component/s: io-java-gcp
• Labels: GCP dataflow pubsubio runner
• Environment: Running locally on MacOS Catalina 10.15.4 and as a Dataflow job in GCP.
• Assignee: Unassigned
• Reporter: Alexander Malyga
• Created: 27/May/20 16:11
• Updated: 04/Jun/22 16:22

Description

My pipeline reads PubSub messages and parses their payload to objects using Gson. I use PubsubIO.readMessagesWithMessageId() to get the PubSub message and the message ID.

I tested the pipeline thoroughly by running it with the DirectRunner in my local machine and everything works fine, but when running it as a Dataflow job in GCP using the DataflowRunner, Gson can't parse the messages properly because the first character of the payload (opening bracket "{") is missing, this only happens when using the DataflowRunner. 

I noticed that the problem no longer happens when using PubsubIO.readStrings() instead of PubsubIO.readMessagesWithMessageId() and getting the payload directly (previously I had to decode the payload using new String(element.getPayload(), StandardCharsets.UTF_8); )