======================================================================= Release Notes for PowerHA SystemMirror Enterprise Edition Version 7.1.3 ======================================================================= These Release Notes contain the latest information about the PowerHA SystemMirror Enterprise Edition software. These release notes discuss: o Introduction o Enhancements and new features included in this release o Documentation Please refer to /usr/es/sbin/cluster/README7.1.3.UPDATE for information about any updates made available via Service Packs. ======================================================================= Introduction ======================================================================= PowerHA SystemMirror Enterprise Edition supports these replication technologies: Software-based mirroring: - PowerHA SystemMirror Enterprise Edition for GLVM (Geographic Logical Volume Manager) Hardware-based mirroring: - PowerHA SystemMirror Enterprise Edition for Metro Mirror - PowerHA SystemMirror Enterprise Edition for SVC Global Mirror - PowerHA SystemMirror Enterprise Edition for IBM DS8700 Global (Asynchronous) Mirror - PowerHA SystemMirror HyperSwap with DS8800 Storage Metro Mirroring - PowerHA SystemMirror Enterprise Edition for IBM XIV Remote Mirroring PowerHA SystemMirror 7.1.3 supports two types of multi site deployments: 1. Stretched Cluster 2. Linked Clusters 1. Stretched Cluster: As the name indicates this involves stretching a single Cluster Aware AIX cluster across a distance. This requires that repository disk be shared across the two sites and as well as the multi cast communication be possible across the sites. Note that the stretched cluster is most suited for short distance multi site deployments. 2. Linked Clusters: Here two separate site specific Cluster Aware clusters are combined using cluster health monitoring links. PowerHA SystemMirror provides the same policy management for workloads across the sites and includes additional options for managing failovers, splits, and merges between sites. Note that linked clusters have a separate repository disk per site and have no requirement for multi cast communications to take place between sites. Linked Clusters are best suited for long distance deployments. ========================================================= o Enhancements and new features included in this release ========================================================= PowerHA SystemMirror HyperSwap with DS8800 Storage Metro Mirroring ------------------------------------------------------------------------------- PowerHA SystemMirror supports the HyperSwap feature with DS8800 metro mirror configuration. HyperSwap is a feature that provides for continuous availability against storage failures. This technology quickly swaps a failed primary storage subsystem with the secondary storage transparently so that applications/middleware do not encounter an outage. The 7.1.3 release introduces the following new features for Hyperswap: o Active-Active Sites allows you to run and active workload at both sites with Hyperswap to protect against storage failures. This feature can be used for Oracle RAC long distance deployments. o Single Node cluster with HyperSwap You can configure a single node cluster with HyperSwap to take advantage of the HyperSwap storage availability, but without the additional availability of a backup node. o Automatic Resynchronization of mirroring when needed o Node level "Unmanaged Mode" Support automatically stops the Hyperswap function when cluster services are stopped with the unmanage option. o Live Partition Mobility is supported when using Hyperswap o Enhanced Repository Disk Swap management - by using Hyperswap to protect the CAA repository disk, the administrator can avoid specifying a standby repository o Dynamic Policy management support - the administrator can modify the HyperSwap policies across the cluster with cluster services active ========================================================================= Split and Merge Handling in PowerHA SystemMirror 7.1.3 Enterprise Edition ========================================================================= In version 7.1.3 the administrator can setup policies which control how the cluster software should respond to a complete split between sites as well as the response when the previously split sites are merging (occurs when the broken network links heal/repaired). The following policies are supported for no sites or stretched clusters: Split: No Action Merge: Majority Split: TieBreaker Merge: TieBreaker The following policies are supported for Linked Clusters: Split: No Action Merge: Priority Split: No Action Merge: Manual Split: TieBreaker Merge: Priority Split: Manual Merge: Manual TieBreaker: a SCSI or iSCSI disk supporting either SCSI-2 or SCSI-3 reservation. When the split or merge happens, the site that obtains the reservation first wins and the nodes at the other site are rebooted. Majority: This policy indicates that the site with the most number of nodes wins and the losing site reboots the nodes when the merge happens. If the number of nodes are the same on both sites, PowerHA SystemMirror chooses the site which reboots based on the node id of the nodes No Action: PowerHA SystemMirror does not take any action when the split happens. Manual: The operator must pick which site continues and which site is to be re-booted. A message is posted on the system console to inform the administrator a choice must be made. The administrator makes the selection by running cl_sm_continue or cl_sm_recover on each site using SMIT, Director, or the clmgr command line utility. The administrator can define the number and frequency of prompts. Priority: Sites are assigned a priority and when a merge occurs, the higher priority site continues. Priority assignment is based on the list of sites defined through SMIT or Director: the first site in the list is higher priority. Split/Merge Policy Updates: Instructions: ---------------------------------------- The administrator must perform the following steps whenever the policies are changed: 1. Make desired changes using SMIT or Director 2. Verify and sync the changes across the cluster Note that Cluster Splits/Merges are should be extremely rare in the case of Stretched Cluster based multi sites. Hence the split/merge policies are most suited for Linked Clusters multi site configurations. ================================ Installation and Migration Notes ================================ Required Software ----------------- 1. All AIX and RSCT requisites of base PowerHA SystemMirror (cluster.es fileset). 2. PowerHA SystemMirror for AIX v.7.1.3 NOTE: In addition to the PowerHA SystemMirror cluster.license fileset, you must install the cluster.xd.license fileset. 3. Install the Metro Mirror filesets for the desired option(s) based on the PowerHA SystemMirror PPRC Management Type. 4. It is highly recommended that you install the latest service for the storage framework fileset devices.common.IBM.storfwork.rte. Storage framework is critical to the storage support infrastructure. ===================================================================== = General Notes for PowerHA SystemMirror for AIX Enterprise Edition = ===================================================================== Setting "fast_fail" for fibre channel attached disk --------------------------------------------------- IBM recommends the customer give careful thought to the setting of the "fc_err_recov" parameter for the fibre channel adapters used to connect to storage managed by PowerHA SystemMirror Enterprise Edition. * The default value of "delayed_fail" will cause extensive retries in response to any problems communicating with the storage subsystem, and delays in the range of tens of minutes before an unrecoverable error is recognized. As noted in the AIX documentation, this value is appropriate if there are no alternative paths to the storage subsystem, or if the storage subsystem holds rootvg, a paging space, or other system critical information. Also, it would be an appropriate choice if the fibre channel link were noisy, or subject to intermittent disruption. * Selecting the value of "fast_fail" causes fewer retries, and more timely recognition by the system of an unrecoverable error. Depending on the nature of the problem, this will allow the AIX MPIO device driver, the AIX LVM, or PowerHA SystemMirror to react more quickly to the problem. This would be appropriate if the reasonable expectation is that the configuration is more likely to experience solid failures - such as severing of the fibre channel links, or loss of power to the storage subsystem - than temporary ones. To change the "fc_err_recov" parameter, it is first necessary to determine the appropriate fibre scsi device or devices. 1. Use the "lspv" command to list all the disks and their owning volume groups known to the system. Note the names of the disk(s) in the volume group(s) on the storage subsystem managed by PowerHA SystemMirror. 2. For each disk, use the command "lsdev -C -l -F parent" to determine the name of the fibre scsi adapter used to communicate with that device. Then, follow the path: smit devices -> FC adapter -> FC SCSI protocol device -> Change/Show characteristics of an FC SCSI device. Select the appropriate fibre scsi device, as determined through the procedure in the above paragraph, and set the "FC Fabric Event Error RECOVERY Policy" to the desired value. Then, set "Apply change to DATABASE only" to "yes". When the command completes, the node will have to be rebooted to make the change effective. (This attribute cannot be changed while the fibre scsi device is in use.) This change can be done from the command line, also. If, for example, the fibre scsi adapter is "fscsi0", then the command would be: chdev -l fscsi0 -a fc_err_recov=fast_fail -P A reboot is still required. =================================== Documentation =================================== IBM no longer uses form numbers for documentation. The books available for PowerHA SystemMirror are listed here. o PowerHA SystemMirror concepts o Planning PowerHA SystemMirror o Installing PowerHA SystemMirror o Administering PowerHA SystemMirror o Troubleshooting PowerHA SystemMirror o Programming client applications for PowerHA SystemMirror o PowerHA SystemMirror for IBM Systems Director o Smart Assists for PowerHA SystemMirror Note that all the smart assist documentation is combined into a single book. o Developing Smart Assist applications for PowerHA SystemMirror o Storage Based High Availability and Disaster Recovery o Geographic Logical Volume Manager PDF and html versions are available online at http://pic.dhe.ibm.com/infocenter/aix/v7r1/index.jsp ------------------------------------------------------ Viewing and installing the documentation files ------------------------------------------------------ You can view the PDF documentation without installing the product. Simply copy the PDFs from the installation media to your desired location or view them directly from the media. If you prefer you can install the documentation by installing the cluster.doc.en_US.es.pdf fileset. Note: Installing all of the documentation requires about 8 MB of space in the /usr filesystem. The documentation (pdf files) is installed in the following directory: /usr/share/man/info/en_US/cluster/HAES