PG replication, high availability, load balancing related clusters, here is a summary for your reference.
PG has the following various replication-based clustering schemes, most of which were not collated at the time. Now there are many of these cluster configuration documents on the Internet, after this document to find the complete point of the centralized backup, do not need to look around every time.
Based on the functionality of the stream replication, the trigger master-slave replication slony, SQL replication-based pgpool-ii, other such as synchronous multi-master replication EDB MMR (this also supports asynchronous replication, played a bit longer, later to find the current record of the summary), asynchronous multi-master replication Bucardo, File system level replication Drdb, shared disk failed to switch NAS later see the situation to fill.
Slony-i Master-slave replication: The primary server asynchronously sends data to the standby server. Based on the trigger, table row-level replication, the primary server may be overhead, not wait for other servers to end the transaction, only asynchronous replication, may lose data, the standby server supports read-only queries, table-level granularity, no conflict, requires the developer to maintain write requests master, read requests to go slave machine???
PGPOOL-II: Copy SQL statement, support multi-master, no primary server overhead, all server completion to end transaction, no data loss, support read-only query from server, library level copy??? ; Conflict Resolution Mechanism required
Stream replication
Copy the Wal log; no primary server overhead; Asynchronous replication does not have to wait for other servers to end transactions; synchronous replication does not lose data; supports read-only queries from the server; instance-level granularity; no conflicts
In addition, Plproxy is a database component developed by Skype. Although plproxy can only do horizontal shards, put it here too. Plproxy is a database component that is developed by Skype. In the process of use, you need to consider issues such as bandwidth and local caching. It restricts the number of agent libraries to be 2 of the time. Proxy implements the forwarding function on a custom function. This requires a large number of business logic to be put on the PostgreSQL server side to complete, reducing the flexibility of the application. It also poses a challenge for possible future database migrations.
Reference:
Http://www.postgresql.org/docs/9.3/static/different-replication-solutions.html
Http://www.postgresql.org/docs/9.3/static/hot-standby.html
Http://www.slony.info/documentation/cmds.html
Http://www.pgpool.net/mediawiki/index.php/Main_Page
Http://plproxy.projects.pgfoundry.org/doc/tutorial.html#_putting_it_all_together
-----------------
Reprint please specify the source:
Blog.csdn.net/beiigang
PostgreSQL Replication Cluster Overview