By Ramakrishnan (Ram) Ramasubramanian & David Graves | Edited by Tim Boles

While migrating Oracle applications from physical to virtual OVM environments can be straightforward, it is important to take into consideration configuration issues regarding the interaction of the Operating System environment to the Virtual Environment. There are some gotchas that can complicate an installation despite assiduous attention to the installation and migration instructions. Figure 1 shows what a typical Oracle VM implementation may look like.

While it is critical to follow the guides that are provided, certain configuration issues must also be taken into account and considered.

Operating System

The version of the guest operating system (OS) is important to consider with migration. The OS from your physical computer should never be migrated to the virtual environment. Always start with a new deployment. The OS you choose for your guest will be determined by the version of the application software (e.g. Oracle E-Business Suite – EBS) you intend to run. While Oracle VM can handle any kind of OS (e.g. Windows, Linux, etc), you should choose one that is supported by Oracle, such as Oracle Enterprise Linux. Having a certified Oracle installation means you can rely fully on Oracle support for both VM and OS issues.

Each version of Linux comes with its own VM architecture: Paravirtualization (PV), Hardware Virtual Machine (HVM), or Hardware Virtual Machine with Para-virtualized device drivers (PVHVM). These settings will drive many configuration changes such as clock synchronization, the maximum number of disks, as well as the device drivers that can be used. For example: Red Hat 5, which runs as a paravirtualized server can have an unlimited number of disks. Red Hat 6, which is typically HVM, can have a maximum number of 5 disks1.

Storage

In designing a virtual environment, storage layouts remain a critical part of the job. One consideration is that swap space is automatically determined by system memory during install and may not reflect the true value required during normal operation 2. Increasing OS slices after layout is possible but adds to quite a bit of complexity. In the worst case may call for the destruction and regeneration of partition tables, a doable, if sweat-inducing operation.

Use Logical Volume Manager (LVM)

In almost every case, deploying Logical Volume Manager (LVM) is the proper choice. The upsides include growing file systems without reboots, aggregating disk slices easily, and easily migrating data between different data sources.

It is even possible to forego the use of partition tables altogether, thus eliminating a layer of complexity as well as removing the need to reboot when you grow a file system. It’s important to remember that without partition tables, commands such as fdisk will report the absence of a file system, so it will be important for you to keep your disk systems properly documented to prevent accidental erasures. Figure 2 is a good example of deploying two file systems (/u01 and /u02) without partition tables. You can see how disks are split up into LUNS which then are presented to the host. The host can either utilize a LUN or pass it through to the LVM of the VM. Each LUN is then represented as a physical extent (PE), and can be grouped into a Volume Group (VG) as a concatenated volume or deploying RAID (this would be redundant since the SAN almost always handles data protection through RAID), and then finally into a Logical Volume (LV). The partition table would be typically installed at this point, after the creation of the logical volume. Instead, you can use make2fs or something similar to create your file system directly on the LV.

Physical LUN connections

Care must be given to the proper design of LUN connections to insure sufficient data capacity. At the very minimum, you need to consider that host bus adapters (HBA) have a maximum data rate. These connections must accommodate migration, backups and application data.

In a physical environment, each application typically has its own server with its own dedicated connections to the SAN. In a virtual environment, many applications may find themselves on the same server. If the data rate of the applications exceeds the capacity of the SAN connection, a situation known as LUN starvation may result, bringing all data transfer to a grinding halt. Figure 3 shows a typical physical configuration while Figure 4 demonstrates that multiple VMs may reside on one server, and possibly flood its connection to the SAN.

Oracle VM installations rely on regular messages between servers to test the health of every server in the system. This heartbeat is used by the system to eject any failed or unresponsive server and restart its VMs on other healthy servers, in a process called fencing.

Fencing becomes important if data rates to the SAN are exceeded. In extreme situations, one or all of the servers may fence if LUN starvation is severe over extended amounts of time.

Resource Configuration

Without a doubt, the limiting resource in a virtualized environment is memory. A distant second is storage access bandwidth. Processing power is almost never a resource limit in practically all situations.

Not only must memory be configured to accommodate all VM’s, but there must also be excess to allow for migrations from physical servers for both maintenance and for High Availability: if you lose a server, you must have sufficient resources to start the VM on another. In the event of a hardware outage if the VM cannot start on another server then it is no better than a physical-based system. The fact that the server is virtualized is of no help if you have no physical server to move it to.

Other considerations for Virtual Environments can be more resilient than physical environments however, the same rules apply. Networks, monitoring, security all continue to play significant parts in the network design. It is also possible to design needlessly complex systems that have the possibility to decrease the ultimate availability of the system. Care must be taken to design as much complexity as needed but no more. Just because you can segregate data, management, and backup traffic into separate VLANs doesn’t mean that you should.

Backups

Backup choices increase in a virtual environment. Typically, the Data LUNS will be backed up using one method while new methods will be available to backup OS LUNS. Depending upon the shared storage systems deployed backups can be traditional or they can take advantage of advanced backup facilities such as EMC Business Continuation Volumes (BCV) present in the storage subsystem. Certain storage systems have the capabilities of integrating with the Oracle system as well 3. It is important to note that snapshots, a familiar feature, is not present in Oracle VM.

Templates

Oracle VM provides a robust templating facility, allowing the quick creation of a template from an existing VM. The use of Oracle VM Templates has also accelerated the deployment of new Oracle Applications Software deployments and provisioning faster by 90 percent. Standard/custom software templates that can be leveraged to quickly and easily roll out different topologies to support customers’ changing IT requirements. Figure 5 shows some of the tools available and how they assist in a standard application lifecycle.

Licensing Costs

With Oracle Virtual Machine implementation support licensing costs are reduced by 25 percent and scalable, reliable IT infrastructure clients demand is affordable 4. Oracle VM enables efficient and cost-effective Oracle Database and Oracle application deployments, with the ability to add pervasive management that drills down to the database and application level.

Performance and Stability

It is critical that we have a high-performance IT infrastructure that can efficiently and reliably accommodate customers’ business requirements. We had success stories wherein Oracle RMAN backups on physical Infrastructure were reporting high I/O Disk reads resolved 80% after migrations to OVM architecture.

Conclusion

Oracle VM is an essential component in the never-ending quest to reduce cost and increase performance through virtualization. While the Oracle VM documentation is thorough and complete, we did encounter peripheral issues that cost us time to undo. Our ultimate goal was eventually achieved with results exceeding our expectations. By taking note of the few gotchas we list here, we hope to make your efforts even more successful and pleasant.

References

1. https://docs.oracle.com/cd/E64076_01/E64077/html/vmrns-limits.html
2. https://access.redhat.com/documentation/en-us/red_hat_enterprise_linux/6/html/storage_administration_guide/ch-swapspace
3. http://www.oracle.com/us/technologies/virtualization/ovm3-storage-connect-459309.pdf
4. http://www.oracle.com/us/media/calculator/vm/vm-home-2132015.html
5. http://www.oracle.com/us/technologies/virtualization/oracle-vm-for-oracle-database-2155841.pdf

About the Author

Ramakrishnan (Ram) Ramasubramanian is a Senior Principal Consultant for NTT DATA Services who has worked in internet technologies for over 15 years. He has hands-on experience in Database technologies, Enterprise Resource Planning architecture as well as cutting edge internet infrastructure technologies. He is an Oracle-certified R12 E-business Suite Professional DBA and an AWS Solutions Associate with lots of field experience in several sectors such as healthcare, manufacturing, retail, telecommunications, printing, and oil and gas. He currently supports clients in need of advanced solutions in ERP, Dev-Ops and Custom applications. He already has popular blogs https://expertappsdba.com in his vertical where he can be reached.

Since graduating from the University of Pennsylvania, David Graves has worked in Internet technologies for over 30 years. He has been Director-level at several billion-dollar companies: At Covad, the first national high-speed internet provider in the US, he oversaw a team of 20 and management of 4 million clients. At Kulicke & Soffa, a USA$800 million technology company, he managed a team of 50 in the global IT group which supported 6 worldwide sites.

He currently runs his own consultancy, providing services to technology, pharma, and media companies in need of advanced solutions in infrastructure and custom applications.