Monitoring, Alerting, and Performance Analysis
This domain tests whether you can read PowerStore Manager telemetry, configure notifications, and diagnose a performance complaint with data rather than guesswork. Expect questions on metric interpretation, alert routing, capacity trending, and node balance.
Monitoring surfaces
- PowerStore Manager provides dashboards, alerts, events, and jobs. Alerts have severities and can be acknowledged; events are the historical record.
- Notification policies route alerts by email, SNMP trap, or syslog. SupportAssist provides Dell remote connectivity and case creation; CloudIQ adds cloud-based analytics but is not required for local monitoring.
- Performance views cover the appliance, individual nodes, volumes, hosts, and file objects, with IOPS, bandwidth, latency, and utilization. Front-end and back-end metrics tell you whether the bottleneck is the host path or the drives.
- Capacity views separate physical used, subscribed, snapshot, and thin savings; set thresholds on physical and snapshot growth.
Reading a performance problem
Start at the widest scope and narrow: appliance latency, then node balance, then the busiest volumes and hosts. A large gap between node utilization usually points to host multipathing or fabric zoning, not to a drive problem. High latency with low utilization suggests queueing or a path issue; high utilization with rising latency suggests a genuine resource limit.
Worked example
Users report slow writes on one volume. The appliance dashboard shows node A at 90% front-end utilization and node B at 35%. The volume is mapped to a host group whose hosts have only one logged-in iSCSI session. The imbalance, not the array's raw capability, is the cause: restore the second path, confirm sessions to both nodes, and recheck node balance before opening a hardware case.
Traps that cost marks
- Sizing or alarming on subscribed capacity instead of physical used plus snapshots.
- Ignoring node imbalance and escalating a path configuration issue as a hardware fault.
- Assuming CloudIQ must be enabled before local alerts work.
- Leaving notification policies unset, so alerts exist only in the UI.