We just tried to deploy our embedded-Flowable (6.3.1) Spring Boot app into a multi-node setup for the first time (previously it was running on a single node, which works fine), and the engine’s DB migration failed because it looks like both instances tried to run in the DB schema migrations at the same time.
Does Flowable use any sort of locking mechanism to manage this scenario? I had been under the impression that it used Liquibase internally, so had assumed that this would occur out-of-the-box, but that doesn’t seem to be the case. Tracing through the code that runs on engine startup (e.g. from the ProcessEngineImpl constructor to the SchemaOperationsProcessEngineBuild command and all the way down to AbstractSqlScriptBasedDbSchemaManager.executeSchemaResource()), I can’t see anything that indicates any sort of attempt to get a DB lock to prevent concurrent schema updates occurring. This code doesn’t even look like it is using Liquibase.
Is there something we need to configure explicitly in order to make Flowable start up safely with multiple nodes sharing the one DB? I couldn’t find anything in the docs about configuring this. I did find some documentation on configuring Liquibase when using forms (e.g. https://www.flowable.org/docs/userguide-form/#database.tables.explained), but we aren’t using forms.
Thanks.