PowerHA Now Replicates To Multiple Targets Sans SAN
May 8, 2023 Alex Woodie
The new release of PowerHA will enable customers to use geographic mirroring to replicate their IBM i data to two or more targets. That one enhancement brings a host of benefits for customers not using SAN-based storage, including new deployment scenarios using on-prem, cloud, and hybrid topologies, as well as lower downtime for hardware upgrades, mirrored Db2 Mirror pairs, and lower RPO for tape backups.
Prior to PowerHA SystemMirror, most IBM midrange customers who were using high availability (HA) software were using solutions based on logical replication, whereby changes to IBM i data are picked up from the local receiver and automatically replicated to the remote journal of the secondary server. Logical replication solutions, such as MIMIX, grew quite mature with a variety of supported topologies, including one server replicating to multiple targets.
The dominance of logical replication began to wane in 2008 with the release of PowerHA, which uses hardware-based replication methods atop IBM’s clustering technology. With PowerHA, all data is moved into independent auxiliary storage pools (iASPs), which are then replicated using two main classes of replication protocols.
Enterprise customers using external storage area networks (SANs) – such as the high-end DS8000 family but later midrange Storwize SANs – had the most options with PowerHA. First and foremost, IBM’s SANs supported LUN-level switching, which essentially enables multiple local servers to have simultaneous access to a single database residing in a SAN, thereby providing HA, but not disaster recovery (DR).
IBM also offered several data replication protocols for the iASPs, including Metro Mirror and Global Mirror (as well as FlashCopy). Enterprise customers could mix and match the LUN-level switching with the replication options to come up with a variety of HA and DR topologies, including single-site, two-site, and even three-site setups. Customers can even sign up with IBM System Lab Services to deploy even more sophisticated setups under the Advanced Copy Services banner, including the famed ACS Metro Global, which can keep three copies of data in synch.
But IBM also offered a simpler PowerHA path for smaller customers and those who just prefer to keep their data residing internally on IBM i direct access storage devices (DASD) as opposed to SANs. For these shops, IBM developed an IBM i-specific replication protocol called geographic mirroring, which allowed PowerHA customers to replicate their iASP data from one IBM i server to another IBM i server.
Geographic mirroring (always in lower case, as opposed to its upper-case Unix-era brethren Global Mirror and Metro Mirror) came in two flavors: synchronous and asynchronous. Just as with the remote journaling replication technology that dominated the logical replication era, whether a customer chose asynchronous or synchronous geographic mirroring depended on how much latency they wanted to introduce in exchange for possibly losing data.
But geographic mirroring setups were always limited to two nodes: a production system and a target. PowerHA customers with SAN setups could mix geographic mirroring into their Metro Mirror, Global Mirror, and LUN-level switching setups, but customers without SANs could not deploy these more complex topologies.
That limitation is going away with IBM PowerHA SystemMirror for i 7.5 TR2, which was announced April 11 with the Technology Refresh (TR) cycle for IBM i 7.5 and 7.4. With the new release, which ships June 23, IBM is now supporting geographic mirroring setups with up to six-nodes, “while keeping the same management simplicity as environments with only two nodes,” it says in the announcement letter.
This opens up all kinds of new possibilities for PowerHA topologies. PowerHA Product Manager Brian Nordland, who works for Fortra, shared some of these scenarios during a session at the recent COMMON POWERUp 2023 conference in Denver, Colorado.
For example, users can replicate their production server to a secondary HA system on prem, while keeping a third server running in the cloud for DR purposes. Or users can replicate from one production system to a backup system on prem, while simultaneously replicating data to two cloud-based systems for HA and DR purposes.
A managed service provider (MSP) could also use the expansion of geographic mirroring to replicate data from a primary IBM i server to an HA and a backup server in one data center, while also replicating data changes to a second three-node IBM i cluster in a second data center, providing significant protection against planned and unplanned downtime.
IBM i shops using Db2 Mirror, which is IBM’s continuous availability solution, can also partake of the new geographic mirroring fun. IBM says that, thanks to the capability to have additional copies of the iASP, customers can now use geographic mirroring to replicate changes from a pair servers linked with Db2 Mirror to a second pair of Db2 Mirror-replicated servers. This setup would nearly eliminate downtime from planned upgrades as well as unplanned outages.
But wait – there’s one more big benefit to this change that will lessen downtime during hardware upgrades. As IBM explains in its announcement letter, PowerHA customers using geographic mirroring previously had to vary off the production iASP when “deconfiguring or reconfiguring” replication.
“With this enhancement,” IBM continues, “the production copy of the IASP can remain online during CFGGEOMIR operations of additional copies of the iASP, significantly reducing the impact on production workloads for hardware migrations.” Users can also change parameters, such as changing from asynchronous to synchronous replication, without varying off, IBM says.
Customers who run backups to tape or VTL from the secondary (or target) node in a PowerHA setup benefit as well. Since the limitation on two nodes is now history, PowerHA customers can spin up a third replicated node to run the backup from, which IBM says will help protect the RPO. In other words, the odds of losing even a little bit of data in the event of an outage effectively goes to nil.
Finally, IBM is also providing a simplified recovery procedure through the capability to failover to a usable back-level copy. IBM delivered this by allowing the system to detach suspended copies of the iASP using the Change ASP Session (CHGASPSSN) common with option *DETACH.
“This enables operations, including easy verification of the back-level copy of the IASP or the ability to perform a save operation from the target, even when replication is suspended without needing to perform a failover,” IBM says in the announcement letter.
IBM also made some changes to the PowerHA Web interface, including new monitoring capabilities. But these pale in comparison to the new options available with geographic mirroring. For PowerHA customers who have been reluctant to move away from internal DASD to the more complex but also more capable world of SANs, the capability to deploy six-node clusters while retaining the simplicity of geographic mirroring will be a boon to availability.
As mentioned, IBM PowerHA SystemMirror for i 7.5 TR2 is slated to ship June 23. You can find more information on PowerHA (although not the new release) at the PowerHA wiki hosted by Fortra at this link.
Big Blue Refreshes The Technology for IBM i 7.4 And 7.5
IBM Puts The Finishing Touches On PowerHA For IBM i 7.5