Answers to Common VCS One Questions
A few weeks back a large number of Product Managers, Technical Product Managers, and Sales Engineers (SEs) gathered for our annual SE Symposium in Las Vegas. VCS One was definitely on everyone's mind, especially with the upcoming 5.0 release. Here's a summary of the Q & A from the VCS One sessions I attended.
Question: How many Policy Masters (PMs) are needed for a VCS One Cluster?
Answer: One PM per cluster. Larger environments will implement multiple VCS One Clusters per data center.
Question: Does the PM have to run in an active/passive cluster?
Answer: Do to the critical role of the PM dedicating hardware is best to ensure performance, reliability, and availability.
Question: Can we run other apps on the PM cluster?
Answer: Not supported in production. Running other apps on the PM could introduce errors/problems that adversely impact PM.
Question: Do we support running the PM inside a virtual machine?
Answer: For test/training yes, but not for production.
Question: What are the minimum requirements for installing a VCS One Cluster?
Answer: A dedicated 2-node PM cluster on RHEL5 or Solaris 10 (Linux recommended/preferred). Refer to the Getting Started Guide for details on supported client Operating Systems.
Question: How many clients are supported in a VCS One Cluster?
Question: Is the 256 client limit physical or virtual?
Answer: 256 includes both physical and virtual servers. This is a soft limit to be used as a guide.
Question: Does VCS One require a dedicated network between the PM and clients?
Answer: Not dedicated. We require a single reliable TCP/IP connection. However, multiple/redundant network connections are always recommended/encouraged.
Question: Is the communication between the PM and client continuous or interrupt driven?
Answer: There are continuous heartbeats between the clients and PM.
Question: Can I control which network the client connects to the PM on?
Answer: Not today.
Question: Does a disconnected client know if he is cut off from the PM?
Answer: Yes, on loss of connection the client stops active monitoring of resources and goes into a disconnected state waiting for the connection to the PM to be re-established. Applications stay running, but are left unprotected as the agents take no action while in a disconnected state. The client continuously retries to connect to the PM. On reconnect the client synchs configuration with PM, agents are told to resume monitoring, resources are probed for any changes that occurred while the client was disconnected, and an updated status is communicated to PM.
Question: Does VCS One use the same agents as classic VCS?
Answer: VCS One and VCS both use the 5.0 agent framework, but not necessarily the same agents. VCS One Agents may include enhancements that are not backward compatible with VCS (i.e. OHR).
Question: How do I install agents in VCS One?
Answer: VCS One agents are included with the quarterly VCS Agent Pack. The agent pack provides an installation tool to install the agents.
Question: What are the best practices for setting load and capacity values in AWM?
Answer: Load and Capacity are static values defined by the admin. Follow KISS principle and don’t over complicate. Users will know how many SGs they want to run on a system and can define numbers to achieve this.
Question: Are the SG priorities set in VCS One static or dynamic?
Answer: Priority is a static attribute but you can use BPA to automate changing priorities based on schedule.
Question: If you kick-out an application how/when does it come back online?
Answer: Apps are moved to a transition state on the Group Transition Queue (GTQ) and will come back online as soon as capacity becomes available.
Question: How does VCS One know when resources become available?
Answer: Clients communicate their status to the PM and when changes occur that take/release resources this is made known to the PM. The PM uses this info together with the GTQ to track/identify when resources become available and bring applications back online.
Question: Explain the “HA” vs. “Start” licensing modes?
Answer: Start entitles user to use VCS One for application start/stop/status and manual switch only. There is no automated failover. HA entitles the user to start/stop/status, switch, and automated failover.
Question: Can we mix “HA” license mode with “start” license mode?
Question: How does VCS One handle permission?
Answer: VCS One has robust role based access and control (RBAC) that allows assigning granular permissions on a per user basis.
Question: Do we keep and audit trail of actions (who did what when)?
Answer: Yes, all actions are logged showing the user that executed the command, the command that was executed, and the time.
Question: Are there any plans to support “non-secure” VCS One Clusters (i.e. clusters without Symantec Product Authentication)?
Answer: No, Symantec Product Authentication is a core component of the VCS One architecture that provides secured/encrypted communication and user authentication.
Question: Do we have a VCS One Simulator? How do I get it?
Answer: Yes, the GA 2.0RU1 is available on Score. Contact PM/TPM for early access to the 5.0 simulator.
Question: Can I copy configurations between the VCS One Simulator and a live VCS One Policy Master?
Answer: Yes, configurations can be saved to XML files and ported between the PM and simulator.
Question: What virtualization technologies does VCS One support?
Answer: Today we support Zones & WPARs. vSphere 4.0 and Solaris LDOM are planned for Q12010. IBM PowerVM and HP IVM will follow. (Note: you can run a VCS One client inside LDOMs/PowerVMs/IVMs today. By virtualization support we are referring to “hypervisor” awareness).
Question: How are Solaris Zones/Projects configured?
Answer: We use Solaris FSS to allocate/set CPU shares as part of zone start/stop. There are additional resources attributes used to control/configure apps that run in zones.
Question: When will VCS One support Hyper-V?
Answer: Mid to late 2010 (prioritized based on customer demand).
Question: When will VCS One support VMware vSphere?
Answer: vSphere 4.0 (console) is planned for Q1 2010. vSphere 4.0i (console-less) support will follow later in the year.
Question: How does VCS One compare to VMware HA and Site Recovery Manager?
Answer: VMware HA and SRM are VMware only solutions designed to protect only VMware VMs. HA and SRM have no visibility to applications running inside the VMs and therefore cannot protect against application failures. VCS One is a tried and proven cross-platform application centric HA/DR solution built on the VCS DNA. VCS One protects physical and virtual servers. For virtual servers we protect both the VMs and the applications running inside them.
Question: Do we replicate the PM as part of DR?
Answer: No. We move apps between different PMs at each site; we do not replicate the PM.
Question: Does VCS One control replication, and what replication technologies are supported?
Answer: VCS One controls replication the same as classic VCS. Check the release notes for a list of replication agents currently qualified.
Question: On DR event, can we run scripts/commands on external systems that are not part of VCS One cluster (i.e. can we update DSN servers)?
Answer: We have agents to update DNS. The Policy Master can run scripts via triggers same as classic VCS.