The target architecture is to:
- Use the Nuxeo Platform built-in clustering mode to ensure HA at Application level;
- Configure PostgreSQL Master/Slave replication mode to ensure HA at Database level.
See the Nuxeo Clustering Configuration page.
Streaming replication allows a standby server to stay more up-to-date than is possible with file-based log shipping.
There are several ways to achieve PGSQL replication: we are just presenting one of the possible option.
Technically, you don't need to replicate the BinaryStore if you use a reliable backend like:
- a NAS that already handles redundancy;
- Amazon S3.
However, if you want to have the full storage replicated in two separated data center, you will want to replicate the BinaryStore too.
Because of the way the BinaryStore is handled (no update, move) , you don't have a lot of constraints:
When the master database is down, the slave must be promoted to master. This means changing changing the
postgresql.conf files and restart. This can be easily scripted.
If your infrastructure provides a virtual IP for the database, the VCS pool (used for the document repository access) is able to reconnect automatically to the database.
Please refer to the Nuxeo backup procedure.