Hello everyone,
I am currently facing the challenge of giving a High Availability solution for a new system we want to work with.
I need PostgreSQL to meet the following requirements:
1. Auto failover in case of node is going down or completely destroyed.
2. Scale-Out architecture because the amount of users continuously going to grow up
3. I must have a Virtual IP referring to the current 'master' node.
4. Full support of PostGIS.
About the environment:
It is virtual servers with RedHat 6.2/6.6 , using pgsql 9.3 or 9.4 .
I have been trying to implement it already with two solution but faced with some serious issues as I will explain :
1. UCARP + repmgr + pgsql 9.3:
after writing the fitting vip-up.sh script I have faced a specific bug with ucarp with after 3 failovers it just started both node as masters and couldn't find anything on the internet that describes such a problem.
2. Redhat Cluster + pgsql 9.3:
a. I have succeed to implement it with shared storage to be managed quite good with redhat cluster. the thing is that I am missing 2 things: active standby and I need it in 'share-nothing' architecture
b. I have followed the following tutorial
https://github.com/smbambling/PGSQL_HA_CLUSTER/wiki/Building-A-Highly-Available-Multi-Node-PostgreSQL-Cluster
but I had no success getting the pgsql service to start on any on the nodes.
Now my question to you lads, has anyone managed to successfully implement such an architecture and how?
Best Regards,
Aviel Buskila