It appears Camunda (per the documentation) supports multitenancy either by (1) shared process engine with single schema with tenantid in each table, or (2) separate process engine per tenant where each tenant data is in a separate schema/database.
From our tests, #2 seems to work fine for handful of tenants in a JVM. However, after a few (> 7) tenants, the total memory footprint of process engines goes high since each process engine consumes nontrivial amount of memory.
We would like to use a single process engine but with separate schema per tenant. This is supported as a 3rd multitenancy option in Activiti and Flowable BPM implementations (via MultiSchemaMultiTenantProcessEngineConfiguration) . Is this possible to do in Camunda? What is the recommended path to achieve this option?