Managing the InfiniBand Fabric
All InfiniBand Switches are discovered automatically during the database machine discovery workflow (see Exadata Database Machine Discovery) and are grouped automatically under the group IBFabric@<switch-name>.
Note:
InfiniBand Fabric target is not available for RoCE Exadata.
- From the Enterprise Manager home page, select Targets, then Oracle Exadata Database Machines and Cloud Services.
- In the Target Navigation pane, select InfiniBand Fabric from the list.
- In the IB Fabric pane, you can view an overview and activity summary for all InfiniBand Switches.
- Click Refresh for an On Demand refresh of the InfiniBand schematic. Updates reflect the real-time data.
The following topics address managing your InfiniBand network:
InfiniBand/RoCE Switch Metrics
- Status / Availability
- Port status
- Vital signs: CPU, Memory, Power, Temperature
- Network interface various data
- Incoming traffic errors, traffic Kb/s and %
- Outgoing traffic errors, traffic Kb/s and %
- Administration and Operational bandwidth Mb/s
The following metrics are available for your InfiniBand Fabric:
Switch Aggregated Status
The Aggregate Sensor takes input from multiple sensors and aggregates the data to identify problems with the switch that require attention. Whenever the sensor trips into an "Asserted" state (indicating a problem) or "Deasserted" (indicating that the problem is cleared) for a component on the switch, associated Enterprise Manager events will be generated.
Response
This is the main metric indicating availability of the InfiniBand/RoCE switch. It is collected every 60 seconds by default through the management interface of the switch.
Switch Configuration
This metric captures the switch configuration. The information collected is valuable only to Oracle Support, which will use it to assist in debugging situations.
Switch Basic Status
This metric gives basic status of the switch like Booted on, Locator light status, Power status and overall status of the switch.
Sensor Status
This metric gives the status of various sensors available in the switch like power supply, fan, motherboard, and cooling.
Switch Port Statistics
This metric provides information on number the of incoming and outgoing errors, incoming and outgoing octets.
Component State
This metric gives the state of various components in the switch like Fan, Motherboard, Power Supply and various InfiniBand and Ethernet ports.
Network Port InfiniBand performance
This metric gives performance data of each InfiniBand port.
Performing Administration Tasks on InfiniBand Networks
Note:
Administrative tasks are not allowed to be performed on RoCE switch.
To perform an administration operation on an InfiniBand Network, follow these steps: