2.4 Time Recommendations

2.4.1 Time Source

The NCS cluster master node uses its server time to record most log entries. To synchronize time for all nodes, you should always use a time source that is external to the cluster. In a virtualized environment, the time source must also be external to host servers because the hypervisor provides virtual clocks to the guest servers. Using a time source that is external to cluster nodes and host servers allows time to be recorded consistently for cluster events.

2.4.2 UTC Time Format

UTC (Coordinated Universal Time) is the standard international time. All time zones are defined as an offset of UTC. UTC does not get adjusted for daylight savings. Use UTC time format instead of local time format on all cluster nodes. In a virtualized environment, ensure that you use UTC time format on the host servers and guests servers.

2.4.3 Tolerance in a Virtualized Environment

The cluster tolerance setting specifies the amount of time the master node gives all other nodes in the cluster to signal that they are alive. In a virtualized environment, a guest server communicates through the hypervisor to the network adapter, which can introduce some latency in communications. If the delay causes problems for the cluster node, you can increase the tolerance from 8 seconds (the default) to 12 seconds. The revised tolerance setting applies to all nodes. If the problem continues, try increasing the tolerance another 4 seconds to 16 seconds.

2.4.4 Tolerance in a Stretch Cluster

The cluster tolerance setting specifies the amount of time the master node gives all other nodes in the cluster to signal that they are alive. In a stretch cluster where nodes are not co-located, the distance between nodes can introduce some latency in communications. If the delay causes nodes to not meet the tolerance settings for the keep-alive signals, you can increase the tolerance for the cluster. Try increasing the tolerance from 8 seconds to 12 seconds. If more time is needed, increase the tolerance another 4 seconds to 16 seconds. Try incremental increases until the time allowed is sufficient under normal network conditions. The revised tolerance setting applies to all nodes.

2.4.5 Quorum and Quorum Timeout

Quorum and Quorum Timeout

2.4.6 Keep Latency in Mind

Keep latency in mind.