The configuration of the mariadb galera cluster does not match the configuration required by camunda officials. As a result, the camunda engine cannot execute the task immediately when processing the concurrent task, and the job generated by SDNCAdapterV1 is suspended until the task's locking time expires.

      Simulation SDNCAdapterV1 USES asynchronous continuation configuration to generate jobs, and USES concurrent gateway plus asynchronous continuation configuration to increase the recurrence probability of job suspension. Test results show that job suspension will occur after about 10 requests.

1、Configuration modification for the mariadb galera cluster with camunda's official requirements

wsrep_causal_reads = 1
wsrep_sync_wait = 7
transaction-isolation=READ-COMMITTED

After 1000 tests of the same BPMN process, no job suspension occurred

2、The mariadb galera cluster was modified to a single node

After 1000 tests of the same BPMN process, no job suspension occurred

Therefore, I believe that the job suspension occurred because the data synchronization mechanism of the database cluster could not meet the requirements of the Camunda engine