Configuring Master agent is to set the number and the location of all Slaves agents. These parameters are assigned when the Master starts, and cannot be changed on the fly.
A particular masters directory is used to configure Master agent. It is located in:
masters subdirectories are agents.
Example, for the service 42, status stable, agent Foo, the following directory structure will be created:
In <agent_name> subdirectory, an empty file for each Slave agent will be created. The name of this empty file is the hostname or IP address of the server hosting Slave agent.
Example, to add two Slave agents hosted on host1 and host2 servers to the previous example:
Changing empty files in the Master directory master is not taken into account immediately and does not stop the Master agent. A restart of the Master agent or of the Query Manager is necessary for the new configuration to be taken into account.
Network connectivity between servers hosting Slaves and Master agents should not be restricted by a firewall. Indeed, ports used by the agents are dynamically allocated, and connections can occur in both directions. In case of communication issues between the Master and Slave agents, Master agent stops automatically after 30 seconds if it could not correctly establish the connection with all configured Slaves.
Choose the number of Slave agents.
To effectively use the resources of Slave agents, reply databases must be tailored to the Master - Slave configuration.
The optimal configuration is achieved when the number of reply database (directories 1/, 2/, and so on) is a multiple of the number of Slave agents, and the size of those reply databases is homogeneous.
It is advised to use the nb_dirs parameter of the afs_search_build filter when running a PAF
The afs_defrag tool can also be used to defragment an existing reply database and reach the desired reply database number.