Intermittent PostgreSQL DDL problems when multiple cluster nodes start simultaneously
Simultaneous startup of multiple PostgreSQL-backed Nuxeo 5.6 cluster nodes seems to cause intermittent PostgreSQL DDL problems. The nodes that report the DDL problems seem to be unstable afterwards. The current workaround is to stage startup of the cluster nodes which is not ideal. Do unit or integration tests currently exist to exercise this specific scenario?
In production noDDL=true should always be used to avoid this.
We don't have immediate plans to have an automatic way to detect concurrent startups, which would need a global lock using a distributed communication system (like Redis).