Cluster Overhead Alarm
The Cluster Overhead alarm is raised when the percentage of time dedicated to cluster-related activities is a high proportion of total active time. The thresholds that represent "a high proportion" are adjustable; to modify thresholds, use the Spotlight Metrics Dialog.
Possible triggers for the Cluster Overhead alarm are:
- A high rate of block transfers between instances in the Oracle RAC cluster. This may happen where there are very hot blocks that cycle throughout instances in the cluster. Typically the Cache Miss Rate Alarm will fire if this is the case. Open the Cluster | Overhead Page to see the segments associated with the relevant wait events.
- High cluster latency. If latency across the interconnect is too high — possibly because the cluster network is misconfigured or inadequate — you may also see the Cluster Latency Alarm. If the Cluster | Latency Page | Average Ping Time chart shows high values, the network itself is the problem.
- High latency in cluster nodes. If individual nodes in the cluster are IO- or CPU-constrained they are slow to respond to requests; often this shows up as GC Buffer Busy waits. You may see the Cluster Latency Alarm, and the overall latency in the Cluster | Latency Page may be poor even though ping times are acceptable.
- Abnormal triggers, such as Oracle bugs or configuration issues. Open the Overhead | Category Events Page to identify problem events.
Possible solutions to this problem include:
- Configuring cluster services (GES processes or the LMS global cache service) to partition workload or the transferred objects themselves (if the problem is a hot block, rather than a hot row).
- Upgrading the interconnect hardware.
- Upgrading one or more nodes in the network.
Cache Miss Rate Alarm