IBM i Global

IBM i 

A space for professionals working with IBM’s integrated OS for Power systems to exchange ideas, ask questions, and share expertise on topics like RPG and COBOL development, application modernization, open source integration, system administration, and business continuity.


#Power


#IBMi
#Power
#Operatingsystems
#Servers
 View Only

Power11 + IBM i Policy-Based HA Campus Cluster: how we cut RTO to minutes (no full IPL) on a 24/7 logistics platform - and what we'd do differently next time

  • 1.  Power11 + IBM i Policy-Based HA Campus Cluster: how we cut RTO to minutes (no full IPL) on a 24/7 logistics platform - and what we'd do differently next time

    Posted 04/30/26 05:06 PM
    Cover image - Campus Cluster topology across Sala A and Sala B with synchronous Policy-Based HA between two FlashSystem 5300 arrays. Reference architecture: 2× Power11 (8c / 128 GB), 2× FS5300 with FCM4, 4× SAN24B-6 32 Gbps, PowerHA SystemMirror for i EE, Safeguarded Copy, BRMS, Copy Services Manager.
    Figure - Cover: Power11 + IBM i Policy-Based HA Campus Cluster.
    If you run an IBM i logistics or distribution platform 24/7, downtime isn't just an SLA discussion - it's a transactional bleed and a customer-experience hit that doesn't recover linearly. This thread is a structured walk-through of an Enterprise Power11 + IBM i Campus Cluster we delivered to a mid-size Italian distribution customer in 7 weeks, with RPO=0 across the campus, RTO measured in minutes, and a Safeguarded Copy logical air-gap protecting the IBM i database against ransomware.
    I'm sharing this because the pattern - Power11 + FlashSystem 5300 with FCM4 + Policy-Based HA + Safeguarded Copy + PowerHA SystemMirror for i EE - is becoming a reference blueprint for IBM i mid-market in Italy, and I want to compare notes with the community on the parts that worked and the parts where we'd refine the approach next time.
    Reference architecture in one line: two Power11 frames (8 cores, 128 GB RAM each), two FS5300 with FCM4, four SAN24B-6 32 Gbps switches, dual VIOS / NPIV, IASP-level failover, immutable Safeguarded Copy + FlashCopy service partition for zero-impact backups.

    1. The business pain - why RTO matters more than RPO on this kind of estate

    On a 24/7 logistics platform supporting web portals and a sales network, the failure mode that hurts most isn't data loss - synchronous replication has effectively solved that. The failure mode that hurts most is the time between the failure and the moment transactions resume.

    • On the legacy Power10 + logical replication + internal disks setup, an IPL on failover was the bottleneck. Even a 10-minute RTO is too long when web sales are flowing through the platform.
    • Customer audit teams in 2026 are now asking for documented RTO numbers, not just SLAs. NIS2 and customer contracts have shifted the conversation.
    • Hardware refresh from Power10 was due, so the question wasn't "do we modernise" - it was "if we modernise, do we redesign HA properly or repeat the past?".

    The answer that fit the 7-week delivery window - and that we'd recommend again - is IASP-level failover orchestrated by PowerHA SystemMirror for i EE on top of synchronous Policy-Based HA between two FS5300 arrays. No full IPL on failover, RPO=0 within the campus, and a clean cyber-vault sitting underneath.

    Horizontal timeline showing five stages from outage to service resumption. T0 outage, T+0 ms PB-HA sync (RPO=0), T+5-15 s PowerHA EE role-swap, T+30-60 s IASP failover (no full IPL), T+1-3 min apps resume on Sala B. Comparison band at the bottom: legacy Power10 with logical replication had RTO ≈ 10+ minutes (full IPL), Power11 with PB-HA + PowerHA EE has RTO ≈ 1-3 minutes and RPO = 0.
    Figure 1 - RPO and RTO budget across the campus: synchronous PB-HA gives RPO=0 between Sala A and Sala B; IASP-level failover keeps RTO in the minute range.

    2. Campus Cluster topology - Sala A and Sala B

    The Campus Cluster spans two computer rooms in the same site (Sala A primary, Sala B secondary). Each side runs:

    • One IBM Power11 frame, 8 physical cores, 128 GB RAM
    • One IBM FlashSystem 5300 with FlashCore Module 4 (FCM4), inline hardware compression and AES-256 encryption
    • Two IBM SAN24B-6 switches at 32 Gbps, dual-fabric A/B, eight FC ports per server (four per fabric, mirrored across nodes)
    • Dual VIOS per frame with SEA failover / LACP on the network side and NPIV on the SAN side, so IBM i LPARs see LUNs and tape directly

    Node A in Sala A is the IBM Production node. Node B in Sala B is the CBU/Test node - fully participating in the failover topology while also serving Test workloads under IBM's Capacity BackUp licensing rules. That's a meaningful licence optimisation that's worth designing in from day 1, not bolting on at the end.

    Architecture diagram of the Campus Cluster across Sala A (Primary) and Sala B (Secondary). Each room hosts a Power11 frame (8 cores, 128 GB RAM, dual VIOS, NPIV, SEA failover) and an FS5300 array with FCM4. The two rooms are connected by PowerHA SystemMirror for i EE at the host level (IASP failover, no full IPL) and Policy-Based HA at the storage level (Consistency Groups for DB2 and tx-log, RPO = 0). Bottom band shows Safeguarded Copy (immutable snapshots, logical air-gap) and FlashCopy service partition (zero-impact daily IASP backups via BRMS and Copy Services Manager).
    Figure 2 - Campus Cluster architecture: Power11 + FS5300 in Sala A and Sala B, dual-fabric 32 Gbps SAN, synchronous Policy-Based HA, Safeguarded Copy and FlashCopy service partition.

    3. Policy-Based HA + PowerHA SystemMirror for i EE - RPO and RTO in practice

    The cluster combines two layers that handle different problems:

    • Policy-Based HA on the FS5300 pair handles synchronous storage replication between Sala A and Sala B. Consistency Groups align the DB2 database and its transaction logs so that the secondary array always has a transactionally consistent image. RPO=0 across the campus.
    • PowerHA SystemMirror for i EE handles the IBM i cluster on top - orchestrating IASP failover, role swap and resource group movement. Crucially, IASP-level failover removes the need for a full IPL when the production role swings to Node B.

    In practical terms: an unplanned outage in Sala A doesn't trigger a 10-minute IPL - it triggers an IASP role swap and applications resume in the minute range. That's the single most important operational change versus the legacy Power10 setup.

    A note on testing: PowerHA EE failover testing should be scheduled and rehearsed quarterly. We did the first three tests during Hypercare; the customer's IT team now owns the playbook and runs them autonomously.

    4. Safeguarded Copy - the immutable cyber-vault layer

    Synchronous replication is great against hardware failure but useless against ransomware: corrupted blocks replicate to the secondary in microseconds. The cyber-vault layer is what makes this architecture cyber-resilient.

    • Safeguarded Copy creates immutable snapshots in a protected domain on the FS5300. The snapshots cannot be deleted or modified - not even by storage administrators. That's a logical air-gap, the right control surface against modern ransomware.
    • The FlashCopy service partition is configured for zero-impact daily IASP backups via BRMS. Backups happen against the snapshot, never against production volumes - no I/O contention, no production slowdown.
    • Recovery flow is rehearsed: pick a clean snapshot, mount it through FlashCopy, validate consistency, promote it. We measured minute-to-second recovery in the rehearsal we did with the customer's audit team observing.

    The combination of Safeguarded Copy + FlashCopy service partition + BRMS gives the customer two practical capabilities the legacy estate didn't have: documented immutability for audit, and a non-destructive way to test recovery without taking production down.

    5. Migration strategy - Side-by-Side, with the legacy Power10 as rollback

    On a 24/7 logistics platform, the migration cutover is the riskiest moment of the whole project. We chose a Side-by-Side approach for one reason: the legacy Power10 environment stays intact throughout the migration, and a simple power-on is the rollback path.

    • Phase 1 - Pilot on Node B. Test LPAR provisioning, end-to-end backup/restore validation, PowerHA EE installation.
    • Phase 2 - Production cut-over weekend. Service stop, GO SAVE 21, LIC/OS install on Power11, data restore, license key update, smoke test.
    • Phase 3 - IASP transformation + PowerHA activation. Library to IASP transformation, Metro Mirror / Policy-Based HA enablement on the FS5300 pair, failover testing.
    • Phase 4 - Safeguarded Copy + FlashCopy service partition activation.
    • Phase 5 - Hypercare, run-book handover, audit documentation.

    The whole sequence took 7 weeks. The legacy Power10 was de-commissioned only after 4 weeks of stable production on Power11 and a successful failover test in Sala B.

    Migration timeline diagram showing five phases over 7 weeks. Phase 1 (Week 1-2) Pilot on Node B with Test LPAR and PowerHA EE install. Phase 2 (Week 3) Cut-over weekend with GO SAVE 21, LIC/OS install on Power11, data restore. Phase 3 (Week 4-5) IASP transformation and PB-HA enablement with failover test. Phase 4 (Week 5-6) Safeguarded Copy and FlashCopy service partition activation. Phase 5 (Week 6-7) Hypercare with run-book handover and audit documentation. The legacy Power10 stays powered-down but ready as a rollback path, and is decommissioned only after 4 weeks of stable production plus successful failover test.
    Figure 3 - Side-by-Side migration timeline: legacy Power10 kept as rollback, 5 phases over 7 weeks, decommissioning only after stable production + failover test.

    6. Three takeaways for IBM i partners and customers

    If you're sizing a similar architecture in 2026:

    • Design IASP-level failover from day 1. Re-platforming an IBM i estate without IASPs into a PowerHA EE cluster later is painful - bake it into the cut-over plan, not the post-go-live wishlist.
    • Treat Safeguarded Copy as a security control, not a backup feature. Wire its snapshot policies and access reviews into the customer's governance, not into the storage team's quiet corner.
    • Side-by-Side migration on Power11 is sober and operationally reassuring on 24/7 estates. Compressing the timeline by killing the legacy environment early is a bad trade - keep the rollback for at least 3-4 weeks of stable production.

    Discussion prompts

    I'd like to compare notes with the IBM i community on three open questions:

    • How are you sizing the FlashCopy service partition for zero-impact IASP backups in a 24/7 estate? Percentage of capacity, fixed reserve, or per-LPAR?
    • For PowerHA EE failover testing - quarterly rehearsals, scripted automation or fully manual? What's worked best for keeping the customer team confident over time?
    • nyone else combining Policy-Based HA with Safeguarded Copy on FS5300/7300/9500 for IBM i workloads? I'm interested in stories where the cyber-vault recovery had to be exercised for real, not as a drill.

    Happy to discuss the parts that didn't go perfectly the first time - and to be challenged on any of these views. The earlier we standardise this pattern across the IBM i partner ecosystem in Europe, the better off our customers will be.

    Roberto Renna · CEO & Lead Solutions Architect · YoctoIT (IBM Platinum Partner)

    #IBMi #IBMPower #Power11 #PowerHA #PolicyBasedHA #IBMFlashSystem #FCM4 #SafeguardedCopy #IASP #IBMChampions



    ------------------------------
    Roberto Renna
    ------------------------------