To monitor the effectiveness of Persistent Memory (PMEM) cache usage with RDMA on Exadata, information on RDMA I/O statistics can be found in: PMEM Cache User Reads section in the AWR Report: This section provides detailed statistics on PMEM read operations, including user reads that leverage RDMA for direct memory access from PMEM. These metrics help gauge the utilization and efficiency of the PMEM cache . Exadata Metrics (LIST METRICCURRENT): By using the LIST METRICCURRENT command within the Exadata environment, real-time metrics related to PMEM usage and RDMA I/O statistics can be accessed. This provides an immediate view of the current performance and activity on Exadata’s persistent memory components . These tools together offer insights into PMEM cache usage and the effectiveness of RDMA for optimized data access.

The two valid reasons are: corruption in the / (root) filesystem . According to 1, this is one of the scenarios that requires a storage server rescue procedure , because it prevents the storage server from booting up normally . The rescue procedure will re-image the root filesystem with a fresh copy of the operating system and restore the configuration files from a backup1. the failure of both physical M.2 disks . According to 2, each Exadata Storage Server X9M-2 has two physical M.2 disks that store the operating system and configuration files for the storage server . If both M.2 disks fail , then the storage server cannot boot up or function properly . The rescue procedure will replace the failed M.2 disks with new ones and re-image them with the operating system and configuration files2.

https://docs.oracle.com/en/engineered-systems/exadata-database-machine/dbmmn/maintainingexadata-storage-servers.html#GUID-AE16A1DA-53C6-4E80-94E5-963AA65373AB

The two commands that must be executed to safely power off the storage server in an Exadata X9M

Database Machine are E and F.

Command E, CellCLI alter GRIDDISK all inactive, will deactivate all of the GRIDDISKS on the affected

storage server. This will ensure that no data is lost during the power-off process.

Command F, shutdown -h now, will shut down the storage server. This will ensure that the storage

server is completely powered off and no data is lost.

For more information on how to properly power off an Exadata Storage Server, refer to the Oracle

Exadata Database Machine X9M Implementation Essentials official text book and resources [1][2].

https://support.oracle.com/knowledge/Oracle%20Database%20Products/1188080_1.html

According to Oracle documentation1, Exadata Storage software has a complete set of automated

operations for hard disk maintenance, when a hard disk has failed or has been flagged as a

problematic disk. It will initiate a rebalance operation to redistribute data across other disks and

mark the failed disk with an amber light.

It collects configuration information automatically (B). OCM gathers detailed configuration

information about a variety of designated Oracle products and components across an enterprise on a

regular basis2.

It collects configuration information on demand ©. OCM can also collect configuration information

manually when requested by the user or by Oracle Support2.

Collected configuration information can be uploaded automatically to Oracle (E). OCM can upload

the collected configuration information to My Oracle Support automatically for analysis and support

purposes3.

Once the physical disk has been replaced, you will need to create a new celldisk and griddisks on the

replaced physical disk. After the griddisks have been dropped from the associated ASM diskgroups,

you can then add the griddisks back into the ASM diskgroup they used to be a member of and

rebalance the data on the effected griddisks. Finally, you should verify that the griddisks located on

the physical disk have been successfully dropped from the associated ASM diskgroups.

According to Oracle.com documents or resources, the four actions that you should take before

proceeding with applying updates to your Exadata Database Machine are:

A . Consult My Oracle Support note 888828.1 to determine the current recommended Exadata

software release123.

B . Check the Exadata Critical Issues My Oracle Support note 1270094.1 for any issues not added to

the latest version of exachk453.

C . Run exachk and resolve only WARNINGS that you have not seen before3.

D . Run the appropriate patchmgr preqequisite check step for each component being updated3.

Statement B is true as the vm_maker —start-domain command can be used to start a virtual

machine manually, as stated in the Oracle Exadata Database Machine X9M Implementation

Essentials Official Text Book (page 65).

Statement E is also true as the vm_maker —auto-start command can be used to configure a virtual

machine to start automatically when the KVM host is started [1][2], as stated in the Oracle Exadata

Database Machine X9M Implementation Essentials Official Text Book (page 66).

According to the Oracle documentation1, two statements that are true about enabling write-back

flash cache are:

Write-back flash cache is on by default for High Capacity Storage Servers in High Redundancy

configurations (A). This is because write-back flash cache provides better performance and reliability

for data file writes than write-through flash cache.

Before write-back flash cache is enabled, you need to drop the Flash Cache first (D). This is because

changing the flash cache mode requires recreating the flash cache with a different configuration.

For Exadata systems, hardware monitoring for components like the motherboard, memory, power,

fans, and network cards on the database nodes is managed through the Integrated Lights Out

Manager (ILOM). The ILOM provides built-in thresholds and sensor readings that are automatically

recognized by Oracle Enterprise Manager.

These thresholds and alerts are already configured within ILOM for Exadata components, ensuring

seamless integration with Enterprise Manager for alerting and monitoring purposes .

Enterprise Manager uses these pre-set ILOM thresholds and does not require additional

configuration for monitoring hardware components. Therefore, administrators benefit from preset

alerting mechanisms that are sufficient for comprehensive hardware status visibility .