Setting the ping-timeout Database Option

To enable failure detection, set the ping-timeout n database option, where n is the timeout interval in seconds. NuoDB recommends an interval timeout of 60 seconds or greater. When network communication with a process times out, the process is evicted from the database by its peers. See Database Options.

The heartbeatPeerTimeout property, in the NUODB_HOME/etc/default.properties file, to an interval of 60 seconds or greater. Brokers and agents ping their connected peer(s) periodically. A peer sends a warning to its agent log if it does not receive acknowledgement of its initial ping after 10 seconds and then continually after that. See Host Properties (default.properties).

The default behavior, heartbeatPeerTimeout=0, is that a peer waits for a ping acknowledgement forever and does not disconnect a non-responsive peer. For a peer to automatically disconnect a non-responsive peer, set heartbeatPeerTimeout to a value greater than the setting of the RaftMaxElectionTimeout property. The default setting for that property is 8300 milliseconds. NuoDB recommends a minimum setting of 20 seconds for the heartbeatPeerTimeout property.

When heartbeatPeerTimeout is set to a value greater than 0 and that length of time elapses before this peer receives a ping acknowledgement from another peer then this peer closes the session and disconnects with the non-responsive peer.

A broker reports disconnected peers to all reachable brokers in the domain. Disconnected peers periodically try to reconnect. A broker tries to peer back into the domain according to the durable domain configuration. An agent tries to reconnect to another broker in the domain.

On this peer’s host, execution of the NuoDB Manager show domain hosts command indicates that the non-responsive peer and any database processes (TEs, SMs, SSMs) associated with the non-responsive peer are UNREACHABLE. The status of a peer as unreachable does not change the durable domain configuration. See NuoDB Manager Show Domain