InfoScale™ 9.0 Virtualization Guide - Solaris

Last Published:
Product(s): InfoScale & Storage Foundation (9.0)
Platform: Solaris
  1. Section I. Overview of InfoScale solutions in Solaris virtualization environments
    1. Overview of InfoScale virtualization solutions
      1.  
        Overview of the InfoScale Virtualization Guide
      2.  
        About Veritas InfoScale support for Solaris virtualization technology
      3.  
        Virtualization use cases addressed by InfoScale
      4.  
        Recommendations for improved resiliency of InfoScale clusters in virtualized environments
  2. Section II. Zones
    1. InfoScale Enterprise Solutions support for Solaris Native Zones
      1.  
        About Solaris Zones
      2. About VCS support for zones
        1.  
          Overview of how VCS works with zones
        2.  
          About the ContainerInfo service group attribute
        3.  
          About the ContainerOpts resource type attribute
        4.  
          About the ResContainerInfo resource type attribute
        5.  
          Zone-aware resources
        6. About the Mount agent
          1.  
            Bringing a Mount resource online in the zone
          2.  
            Setting the attribute values for a Mount resource for NFS mounts
          3.  
            Configuring a direct mount of VxFS file system in a non-global zone with VCS
        7.  
          About networking agents
        8.  
          About the Zone agent
      3. Configuring VCS in zones
        1. Prerequisites for configuring VCS in zones
          1.  
            Method for file system access inside non-global zone
          2.  
            Using custom agents in zones
        2. Deciding on the zone root location
          1.  
            Creating a zone with root on local disk
          2.  
            Creating a zone with root on shared storage
        3.  
          Performing the initial internal zone configuration
        4.  
          About installing applications in a zone
        5. Configuring the service group for the application
          1.  
            Resource dependency diagrams: zone root on local disks
          2.  
            Resource dependency diagrams: zone root on shared disks
        6.  
          Configuring a zone resource in a failover service group with the hazonesetup utility
        7.  
          Configuring zone resource in a parallel service group with the hazonesetup utility
        8.  
          Configuring multiple zone resources using same VCS user for password less communication
        9.  
          Modifying the service group configuration
        10.  
          Verifying the zone configuration
        11.  
          Synchronizing the zone configuration across cluster nodes
        12.  
          Performing maintenance tasks
        13.  
          Troubleshooting zones
      4.  
        Adding VxFS file systems to a non-global zone
      5.  
        Mounting VxFS as lofs into a non-global zone
      6.  
        Mounting VxFS directly into a non-global zone from global zone
      7.  
        Mounting VxFS as VxFS inside a non-global zone
      8.  
        Adding a direct mount to a zone's configuration
      9.  
        Benefits of a VxFS mount in a non-global zone over VxFS mount from global zone into the non-global zone
      10.  
        Cluster FileSystem mounts
      11.  
        Concurrent I/O access in non-global zones
      12.  
        Veritas Extension for Oracle Disk Manager
      13. Exporting VxVM volumes to a non-global zone
        1.  
          VxVM devices in Oracle Solaris global zones
        2.  
          Removing a VxVM volume from a non-global zone
      14. About InfoScale SFRAC component support for Oracle RAC in a zone environment
        1.  
          Supported configuration
        2. Known issues with supporting a InfoScale SFRAC component in a zone environment
          1.  
            CFS mount agent does not support mounting VxVM devices inside non-global zones
          2.  
            Issue with VCS agents
          3.  
            Stopping non-global zones configured with direct-mount file systems from outside VCS causes the corresponding zone resource to fault or go offline
          4.  
            Error message displayed for PrivNIC resource if zone is not running
          5.  
            Warning messages displayed when VCS restarts
          6.  
            The installer log of non-global zone contains warning messages
          7.  
            Issue with CFS mounts
      15.  
        Configuring Solaris non-global zones for disaster recovery
      16. Software limitations of InfoScale support of non-global zones
        1.  
          Administration commands are not supported in non-global zone
        2.  
          VxFS file system is not supported as the root of a non-global zone
        3.  
          QIO and CQIO are not supported
        4.  
          Package installation in non-global zones
        5.  
          Package removal with non-global zone configurations
        6.  
          Root volume cannot be added to non-global zones
        7.  
          Some Arctera Volume Manager operations can cause volume device names to go out of sync
        8.  
          Solaris branded zone support
  3. Section III. Oracle VM Server for SPARC
    1. InfoScale Enterprise Solutions support for Oracle VM Server for SPARC
      1.  
        About Oracle VM Server for SPARC
      2.  
        Terminology for Oracle VM Server for SPARC
      3. Oracle VM Server for SPARC deployment models
        1.  
          Split InfoScale stack
        2.  
          Guest-based InfoScale stack
        3.  
          Layered InfoScale stack
      4. Benefits of deploying Arctera InfoScale Enterprise solutions in Oracle VM server for SPARC
        1.  
          Standardization of tools
        2.  
          Array migration
        3.  
          Moving storage between physical and virtual environments
        4.  
          Boot Image Management
      5. Features
        1. InfoScale features
          1. The vxloadm utility enables access to a file system contained in a VxVM volume from the Control Domain
            1.  
              Examples of using the vxloadm utility
          2. The vxformat utility automatically relabels the virtual disk backed by a VxVM volume in the guest domain
            1.  
              Example of using the vxformat utility
        2. Oracle VM Server for SPARC features
          1.  
            Guest domain migration
          2.  
            Virtual I/O dynamic reconfiguration
      6. Split InfoScale stack model
        1.  
          How Arctera InfoScale Enterprise works in the Oracle VM Server for SPARC
        2.  
          InfoScale features restrictions
      7. Guest-based InfoScale stack model
        1.  
          How InfoScale Enterprise Solutions works in the guest domains
        2. About InfoScale Enterprise in an Oracle VM Server for SPARC environment
          1.  
            Supported configurations with Arctera InfoScale Enterprise and multiple I/O Domains
          2.  
            Arctera InfoScale Enterprise limitations
        3. InfoScale features restrictions
          1.  
            Arctera Volume Replicator bunker replication
          2.  
            Mirroring across controllers using vxassist the mirror=ctlr option
          3.  
            DMP SCSI bypass
          4.  
            Event Source Daemon (vxesd) fabric monitoring capabilities
          5.  
            Physical WWN for a path
      8. Layered InfoScale stack model
        1. DMP as a multi-pathing solution in an Oracle VM Server for SPARC environment
          1.  
            Enabling DMP in the control and alternate I/O domains
          2.  
            Enabling DMP path failover in the guest domain
          3.  
            Enabling I/O fencing in the guest domain
        2.  
          How InfoScale Enterprise Solutions works in the Oracle VM Server for SPARC
        3.  
          InfoScale features restrictions
      9. System requirements
        1.  
          Hardware requirements
      10.  
        Product release notes
      11.  
        Product licensing
      12. Installing InfoScale in a Oracle VM Server for SPARC environment
        1.  
          Installing and configuring Oracle VM Server for SPARC and domains
        2. Installing InfoScale in the control domain or guest
          1.  
            Installing the split InfoScale stack model
          2.  
            Installing the guest-based InfoScale stack model
        3.  
          Installing Arctera File System in the guest domain
        4.  
          Verifying the configuration
      13.  
        Exporting a VeritasArctera volume to a guest domain from the control domain
      14. Provisioning storage for a guest domain
        1.  
          Provisioning Arctera Volume Manager volumes as data disks for guest domains
        2. Provisioning Arctera Volume Manager volumes as boot disks for guest domains
          1. Using VxVM mirrored volumes as boot devices for LDOMs
            1.  
              Provisioning boot disks for a guest domain
            2.  
              Using VxVM snapshot as a backup copy of the boot image during the upgrade
            3.  
              Restore the original environment
      15.  
        Using Arctera Volume Manager snapshots for cloning logical domain boot disks
      16.  
        Support of live migration for Solaris LDOMs with fencing configured in DMP mode
      17.  
        Configuring Oracle VM Server for SPARC guest domains for disaster recovery
      18. Software limitations
        1.  
          When an I/O domain fails, the vxdisk scandisks or vxdctl enable command take a long time to complete (2791127)
        2.  
          Resizing a Arctera Volume Manager volume (exported as a slice or full disk) does not dynamically reflect the new size of the volume in the guest
      19. Known issues
        1. Guest-based known issues
          1.  
            Guest domain node shows only one PGR key instead of two after rejecting the other node in the cluster
          2.  
            Disk paths intermittently go offline while performing I/O on a mirrored volume
        2. Split InfoScale stack known issues
          1.  
            Caching of data writes on the backend volume in the service domain
          2.  
            A volume can become inaccessible from the guest in the event of control domain reboot
    2. Cluster Server support for using CVM with multiple nodes in a Oracle VM Server for SPARC environment
      1.  
        Clustering using Cluster Volume Manager
      2. Installing InfoScale on multiple nodes in a Logical Domain
        1. Reconfiguring the clustering agents for Cluster Volume Manager
          1.  
            Removing the vxfsckd resource
          2.  
            Creating CVMVolDg in a group
      3. Cluster Volume Manager in the control domain for providing high availability
        1.  
          Provisioning storage to guests with Flexible Storage Sharing volumes of control domain
    3. VCS: Configuring Oracle VM Server for SPARC for high availability
      1. About VCS in a Oracle VM Server for SPARC environment
        1.  
          Benefits of using VCS in Oracle VM server for SPARC environments to manage logical domains
        2.  
          Dynamic reconfiguration of memory and CPU of a guest domain
        3.  
          AdaptiveHA for logical domains
        4.  
          Cluster Server requirements
        5.  
          Cluster Server limitations
        6. Cluster Server known issues
          1.  
            Shutting down the control domain may cause the guest domain to crash (1631762)
      2. About Cluster Server configuration models in an Oracle VM Server for SPARC environment
        1. Cluster Server setup to fail over a logical domain on a failure of logical domain
          1. Configuration scenarios
            1.  
              Network configuration
            2. Storage configurations
              1.  
                Veritas Volume Manager exposed volumes
              2.  
                Image files
            3.  
              Configuring logical domain
            4.  
              Installing Cluster Server inside the control domain
          2.  
            Creating the VCS service groups for logical domains
          3. Verifying a logical domain service group failover
            1.  
              Using hagrp -switch command
            2.  
              Other verification scenarios
        2. Cluster Server setup to fail over an Application running inside logical domain on a failure of Application
          1.  
            Configuring VCS to fail over an application on a failure
      3. Oracle VM Server for SPARC guest domain migration in VCS environment
        1.  
          Overview of a warm migration
        2. Overview of a live migration
          1. VCS initiated migration of service groups that have LDOM resource configured to monitor Oracle VM for SPARC guest domains
            1.  
              Troubleshooting migration-related issues
          2.  
            User initiated migration of Oracle VM guest domains managed by VCS
        3.  
          Prerequisites before you perform domain migration
        4.  
          Supported deployment models for Oracle VM Server for SPARC domain migration with VCS
        5.  
          Migrating Oracle VM guest when VCS is installed in the control domain that manages the guest domain
        6.  
          Migrating Oracle VM guest when VCS is installed in the control domain and single-node VCS is installed inside the guest domain to monitor applications inside the guest domain
        7.  
          Migrating Oracle VM guest when VCS cluster is installed in the guest domains to manage applications for Oracle VM Server for SPARC version 2.1 and above
        8.  
          Migrating Oracle VM guest when VCS cluster is installed in the guest domains to manage applications for Oracle VM Server for SPARC version 2.0
      4. About configuring VCS for Oracle VM Server for SPARC with multiple I/O domains
        1.  
          About Alternate I/O domain
        2.  
          Setting up the Alternate I/O domain
        3.  
          Configuring VCS to manage a Logical Domain with multiple I/O domains
      5. Configuring VCS to manage a Logical Domain using services from multiple I/O domains
        1.  
          A typical setup for a Logical Domain with multiple I/O services
        2.  
          Identify supported storage and network services
        3.  
          Determine the number of nodes to form VCS cluster
        4.  
          Install and configure VCS inside the control domain and alternate I/O domain
        5. Configuring storage services
          1.  
            Configuring virtual disk with LUN, Shared Disk, or CVM Volume
          2.  
            Virtual disk configurations with DMP
          3.  
            Configuring storage services when back-end device is a ZFS volume
        6.  
          Configure storage service groups
        7.  
          Configure network service groups
        8. Configure a service group to monitor services from multiple I/O domains
          1.  
            Type of service group configuration for the AlternateIO resource
        9. Configure the AlternateIO resource
          1.  
            Sample service group configuration for the AlternateIO resource
        10. Configure the service group for a Logical Domain
          1.  
            Sample configuration for LDOM service group
        11.  
          Failover scenarios
        12.  
          Recommendations while configuring VCS and Oracle VM Server for SPARC with multiple I/O domains
        13.  
          Sample VCS configuration for AlternateIO resource configured as a fail over type
      6.  
        Configuring VCS on logical domains to manage applications using services from multiple I/O domains
    4. SFRAC support for Oracle VM Server for SPARC environments
      1.  
        About deploying SFRAC in Oracle VM Server for SPARC environments
      2.  
        Sample configuration scenarios
      3.  
        Preparing to deploy SFRAC in logical domain environments
      4.  
        SFRAC with Oracle RAC database on I/O domains of two hosts
      5.  
        SFRAC with Oracle RAC database on guest domains of two hosts
      6.  
        SFRAC with Oracle RAC database on guest domains of single host
      7.  
        SFRAC with Oracle RAC database on I/O domain and guest domain of single host
    5. Support for live migration in FSS environments
      1.  
        About live migration in Flexible Storage Sharing (FSS) environments
      2.  
        Performing live migration of Oracle VM Server for SPARC systems in Flexible Storage Sharing (FSS) environments
    6. Using SmartIO in the virtualized environment
      1.  
        About SmartIO in the Solaris virtualized environments
      2.  
        Performing live migration between LDOMs in the SmartIO environment
  4. Section IV. Reference
    1. Appendix A. Where to find more information
      1.  
        Arctera InfoScale documentation
      2.  
        Solaris virtualization documentation
      3.  
        Service and support
      4.  
        About Services and Operations Readiness Tools (SORT)

Shutting down the control domain may cause the guest domain to crash (1631762)

Set up

Two Oracle SPARC Enterprise T5240 server physical boxes, each with a control domain and a guest domain. The guest domains in each of the physical boxes form a two node cluster. The nodes are named node 0 and node 1 in the following text.

Symptom

Gracefully shutting down the control domain of node 0 causes the guest domain of node 0 to crash.

Analysis

Even though the guest domain can continue to function when the control domain is shut down, the heartbeats between node 0 and node 1 are lost as the control domain shuts down. As a result, the cluster forms two separate sub-clusters without the sub-clusters being able to see each others' heartbeats.

I/O fencing resolves the split brain situation and determines that only one sub-cluster will continue to function while the other sub-cluster should panic. Therefore, the panic of node 0 is expected behavior.

Resolution

None; this is expected behavior. However, Arctera recommends keeping the control domain highly available for the proper function of the InfoScale Enterprise and InfoScale for Oracle RAC stack in the guest domains.

If you have set up a virtual private LLT heartbeats between the two guests (node 0 and node1), the guest will not crash.