Applicable toApplications Manager Versions: v7.1, v8.0
Additional Questions and Answers
Q: What is failover?
A: The term "failover" can mean different things, so the goal of this knowledge base entry is to clearly define what failover means when working with Applications Manager.
Here are a few customer definitions of failover:
- Applications Manager is installed on a virtual layer that is installed on top of multiple physical nodes. With a virtual layer, if one node or machine fails, the Software installed on it (in this case, Applications Manager) will be unaffected and continue to run on the remaining nodes. From an Applications Manager perspective, the Master would not know multiple physical nodes are involved, and the configuraiton would be tailored for only one environment.
- In the event of a Master failure, a backup Master can be manually or automatically started. This is sometimes referred to as a parallel, mirror, or clone solution, and it usually involves clustering or VPS*.
- A machine with multiple network cards for network fault tolerance. If one card fails, Applications Manager communication can "fail over" and continue using the remaining card.
Please note that Microsoft Windows Load Balancing Service is not a failover tool for Applications Manager, because it directs traffic to other Web servers. It is a networking tool rather than a Web tool.
Q: When a Master is on a cluster, how many licenses does that require?
A: It depends on your configuration. This is a question that you should ask your Automic Sales Account Manager.
Q: Does Applications Manager have better performance on a cluster?
A: No, Applications Manager works with clusters only for failover purposes, not increase performance. This means that the Applications Manager Master runs on only one machine and can run on another machine in a failover situation.
Q: How does the Applications Manager Master look at the grid with Oracle 9i or 10g?
A: In Oracle 9i and 10g, the grid is transparent to Applications Manager and Applications Manager has only one active Master at any one time. We recommend using a grid for failover only, not for load balancing. Using the gird for load balancing with Applications Manager can cause performance problems.
Q: I'm interested in a failover solution that is developed, implemented, and supported by Applications Manager. What should I do?
A: It depends. As a possible fault tolerant solution you can set up two Agents on two machines that have the same supporting applications. Then you can create an Applications Manager Agent group that includes both Agents, and make sure the Agent group is NOT set to use Mulit-Execution (this will cause the jobs to run on all Agents in the group). Then if one Agent in the Agent group is unavailable (for example, if the agent's machine crashes), then Applications Manager will run new jobs on the Agent in that Agent group that is available. Keep in mind that if there was a machine failure, the jobs already running on that machine woujld stop and Applications Manager would report an error. for more information, contact Automic support.
If you need assistance with implementing a customer failover solution developed specifically for your site, please contact our consulting group at GSO@automic.com.
* A cluster is one IP for many machines, and VPS is one machine with many IP's (an IP for each segment, or virtual private server). These configurations might have the database in the cluster or VPS, or the database might be outside.
A cluster configuration might involve two physical machines that have the same virtual IP to the rest of the network. Both machines might have a copy of the same Master, but only one Master would be running at any time. If one physical machine fails, the Master on the other machine can be started.
A VPS configuration might have one machine that has two individual segments, each segment with it's own IP. Each segment would view the other segments on the same machine as a different physical machine. If the physical machine fails (such as a power loss), all virtual segments fail. If one of the segments fails, the other segments are not affected.