My work focused on a high-availability environment where customers maintained two or three data centers designed for disaster recovery solutions. I managed local clusters as well as global clusters, and when a service failed in a particular environment, it automatically moved to a different region. The entire solution was designed with InfoScale at its core.

I find InfoScale's automated Stack Aware Recovery feature to be very beneficial when recovering from ransomware events. The system automatically moved services to alternate nodes, behavior I observed while running on UNIX systems that had underlying issues. Veritas was very active and proactive, notifying me in advance about cluster conditions and recommended actions. From a recovery perspective, my main goal was to establish application recovery, which is why I selected this solution. Some critical machines in the environment still run Veritas because it excels at recovery, is easy to understand, and facilitates root cause analysis. Most of the root cause analyses I performed over ten years were not related to Veritas. I consider it a great product that meets expectations. The primary concern is licensing cost, as the customer is unwilling to invest further and has begun cost-cutting measures. With cloud adoption, they are moving workloads to the cloud, believing it offers greater benefits than on-premises solutions. All VCS instances and Veritas clusters ran on-premises only, with nothing moved to the cloud. Most licenses have already expired, and the customer has allowed me to continue using them while exploring alternative solutions. The application team is redesigning applications from scratch, with several already migrated to microservice architecture in Kubernetes in the cloud.

From a recovery standpoint, InfoScale is excellent and easy to manage. A single technical person can handle 100 machines or one application spanning multiple clusters. I utilized the Virtual Business Services feature to design the solution, enabling all databases, frontend machines, backend machines, and related components to move to different clusters seamlessly without any issues.

InfoScale has significantly reduced downtime for my customer. I encountered unusual split-brain issues. Because I did not utilize all cluster features such as I/O Fencing, which requires additional setup and licensing cost, the solution was not designed with I/O Fencing. When split-brain occurred, I had to investigate the cause, protect the data, and determine remediation steps. For data protection, I implemented SCSI-3 Persistent Reserve at the storage level instead of using I/O Fencing.

What I value most about InfoScale is its ease of use and clear visibility into environment operations, particularly in large environments. I set up Veritas Operation Manager, which interacts with all clusters and provides a central management location for the entire clustering environment. The Cluster Manager graphical interface tool is excellent for identifying problems easily. The solution is robust and rarely causes issues. When cluster problems arise, I can identify that the actual problem exists elsewhere, with Veritas alerting me that external factors are affecting cluster behavior. This makes identifying root causes straightforward in a solid environment.

Comparing InfoScale to HACMP (now called PowerHA by IBM), InfoScale is significantly easier. Before 2010 and 2011, my customer used PowerHA before transitioning to InfoScale. We started with version 5 and progressed to version 7.3. The last running instances are version 7.2 or 7.3. We had strong confidence in the product, and architects were very satisfied with its performance matching our requirements exactly. Pricing is the only issue, as the customer cannot justify additional investment and is phasing out instances in favor of custom solutions.

I have used the application-aware failover feature. The environment runs very few single-node systems with application HA, which continue to function but are not actively used. The customer is working to remove applications from Veritas control. This feature remains intact and operational but has no upgrade or evolution plans.

Regarding layering dependencies across web, application, and data tiers, I worked with the Virtual Business Service feature within Veritas Operation Manager. I created Virtual Business Services with dependencies for frontend web servers, databases such as DB2 or Oracle, and defined startup sequences where the database starts first, followed by frontend services only after the database is live and running. This layering system was quite helpful, requiring only a single click to trigger the entire process, leaving Veritas to manage everything automatically.

Beyond pricing, there are areas where I would like to see InfoScale improved or enhanced. Veritas offers three management approaches. The first, which Veritas currently recommends, is Veritas Operation Manager. The second is the Cluster Manager Java Console graphical interface. The Cluster Manager Java Console has not been revised since version 6.1 or 6.2. This tool was critical for me, particularly valuable when managing small cluster footprints of 20 to 30 server nodes. I relied heavily on this tool, but Veritas has moved away from it in favor of Operation Manager. I recommend Veritas continue evolving this tool rather than discarding it. The third approach is the command line, suitable for individuals with extensive Veritas expertise and experience, but command line use in live environments consumed excessive time, leading me to prefer the graphical interface.

Apart from pricing, I have not discovered disadvantages. The product is excellent. My concern is Veritas discarding the Cluster Manager Java Console in favor of Veritas Operation Manager. Setting up Operation Manager requires time and a dedicated server that runs continuously. I had to create a single server just for Veritas Operation Manager. While this works well for larger environments with hundreds of clusters, it is less useful for smaller deployments. I still recommend Veritas reconsider this application and evolve it by incorporating new features from Veritas Operation Manager. Adding these new features to the Java console would be beneficial because that tool runs on my laptop without consuming environment resources, and I can connect directly to clusters from my laptop. I am not opposing Veritas Operation Manager, which is excellent and resembles hardware management consoles for power machines, but smaller tools that previously performed these tasks should remain as options to provide clients with greater ease.

From a features and functionality perspective, I do not find missing features in InfoScale at this moment. However, I am not actively using Veritas, managing only legacy machines on older hardware. I am upgrading operating systems but not Veritas due to contract expiration and end-of-life status. The contract is not being renewed because the customer wants to move away. Since I have not logged into VCS since 2021 and transferred responsibilities to another team, I am unaware of features arriving in version 8 or beyond and cannot comment specifically on recent Veritas introductions.

I have used InfoScale for about eight years.

Regarding stability and scalability, I have not experienced performance issues or limitations. Adding nodes is straightforward. The maximum cluster I managed ran five nodes, and scaling to six nodes was easy. I created a machine, deployed Veritas guidelines, joined it to the cluster membership, and continued from there. Management through the graphical interface or command line is straightforward. Veritas supports a maximum of 32 nodes, though I never exhausted that capacity.

When I encounter situations I cannot resolve or understand, or when incidents require vendor input or investigation, I contact Veritas customer service. I raise tickets, and they participate in root cause analysis and incident fixes. This interaction is limited because the product is stable and robust, rarely causing problems. Once the Veritas InfoScale contract expired, my customer designed alternative solutions outside InfoScale. They began phasing out and decommissioning InfoScale environments, reducing from 500 cluster nodes to approximately 50 nodes. Product interaction with the support team is now very limited.

The Cluster Manager graphical interface was excellent for identifying problems easily. The solution was robust and rarely caused issues. When cluster problems arose, I identified that the actual problem existed elsewhere, with Veritas alerting me that external factors affected cluster behavior, making root cause identification straightforward in a solid environment. HACMP, now called PowerHA and developed by IBM, is significantly harder. In my early environment before 2010 and 2011, my customer used PowerHA before transitioning to InfoScale. We started with version 5 and progressed to version 7.3.

Comparing InfoScale to other clustering products, the heartbeat implementation stands out. HACMP does not run a heartbeat; they now have a setup running heartbeat on disk. In VCS, I maintained three different types of heartbeats. If one failed, another remained active. When the second failed, the third provided redundancy. Another excellent Veritas feature is the ability to freeze applications or service groups. Whenever maintenance was scheduled or for other reasons, I could freeze them. This option does not exist in any other solution, making it outstanding. The freeze option is exceptional in Veritas.

I was involved in the deployment process and initial setup of InfoScale. We designed the solution, not only myself but also an architect. My role as an SRE involves running the service and maintaining uptime rather than participating in design and solutioning. In the initial days, I set up cluster servers from scratch, installing all LPARs on AIX and configuring storage, then loading Veritas. One important Veritas feature I should mention is that it has maintained hybrid solutions since 2006 or 2007. Veritas can run two different cluster nodes with two different operating systems, such as one Windows node and one Linux node. This feature does not exist in any other solution and is impossible elsewhere. Veritas offers this capability, which I appreciate. I never utilized this feature, but its availability demonstrates Veritas' comprehensive approach.

Comparing InfoScale to other clustering products I have worked with, the heartbeat implementation is distinctive. HACMP does not run a heartbeat; their current setup runs heartbeat on disk. In VCS, I maintained three different types of heartbeats, providing redundancy if one failed. Another excellent Veritas feature is the ability to freeze applications or service groups. Whenever maintenance approached or for other reasons, I could freeze them. This option does not exist in any other solution, making it exceptional. The freeze option is outstanding in Veritas.

I recommend InfoScale to others based on my extensive experience. Previously, when advising another customer interested in IBM solutions such as PowerHA and HACMP who had purchased an IBM solution, I suggested they select Veritas instead. Despite being an IBM employee at that time, I recommended they not purchase PowerHA and proceed with Veritas. They discussed my explanation and decided to trust my assessment, reasoning that managing two different systems would be complicated and that my comfort with Veritas made it the better choice.