Systems with Master, Backup, and Slave Index Servers
You can enhance master/slave systems by adding backup servers (backup index servers and backup queue servers). Such enhanced systems offer additional high availability for indexing.
Each master index server manages some of the indexes. If a master index server goes down, indexing does not normally take place for affected indexes. You implement backup index servers to avoid this. A backup index server can replace a master index server if it becomes unavailable. The backup index server is inactive if the master index server is available.
The same is true for the queue server: If a master queue server goes down, queuing normally does not take place for the documents affected, and this means that indexing cannot take place either. You implement backup queue servers to avoid this. A backup queue server can replace a master queue server if it becomes unavailable. The backup queue server is inactive if the master queue server is available.
The TREX data has to be stored centrally in systems with backup servers. Otherwise the master and backup servers cannot access the data.
You can build systems with backup servers in the following way:
- One backup server per master server
- One backup server for all master servers
The following factors dictate which variant to choose.
- The number of master servers
- The number of master servers that are expected to be down for maintenance at the same time
The smallest recommended system with backup servers consists of one file server, one backup server, one master server, and two slave servers.
The graphic below depicts this system.
As many index servers must run on the backup host as run on the master host. If two index servers run on the master host, two index servers must run on the backup host.
The graphic below depicts a larger system with multiple master and backup hosts.
In this system, both master hosts can be down at the same time, because each has its own backup host.
One Backup Server for All Master Servers
You can build systems in which one backup host is assigned to all master hosts. Only one master host can be down at any one time in such systems. If multiple master hosts with a full load go down, one backup host cannot take on the entire load.
The graphic below depicts a system with two master hosts that share a backup host.
