IBM FlashSystem

 View Only

 Flashsystem PBHA and PBR 4 system (2 site)

Jump to  Best Answer
NGUYEN HIEU's profile image
NGUYEN HIEU posted 08/31/26 12:09 AM

Hi there,

I'm planing HA and Replicate as:

DC site:

  • Components: Two FlashSystem 9600 units (System A and System B).
  • Configuration: These two systems are paired within a single Storage Partition using the Policy-Based HA (PBHA) policy.
  • Data block: This HA partition contains 3 Volume Group (VG1, VG2 and VG3). Data is written synchronously (Active-Active) to both storage units A and B.

DR site:

  • Components: Two FlashSystem 9600 units (System C and System D).
  • Use Policy-Based Replication (PBR) assigned individually to each Volume Group.
  • PBR stream 1: VG1, VG2 is replicated via the WAN link form DC to the system C at the DR site.
  • PBR stream 2: VG3 is replicated via the WAN link form DC to the system D at the DR site.

Please see attached picture for more info.

My Question: Does IBM Storage Virtualize support installation and configuration based on the above model?

omar tajioue's profile image
omar tajioue  Best Answer
Hi,
Short answer: not exactly as drawn — but yes, it is achievable with one adjustment.
The limitation: a single storage partition can only have one DR relationship to one DR system. From one PBHA partition, you cannot replicate some volume groups to System C and others to System D. IBM Development confirmed this: DR volume groups within a single partition must share the same DR target.
The supported design: split your workload into two storage partitions on the same A/B PBHA pair:
 
Partition 1 → VG1 + VG2 → PBHA between A and B → PBR to System C
Partition 2 → VG3 → PBHA between A and B → PBR to System D
 
Both partitions run active-active HA across the same two production systems (A/B), and each partition has its own independent DR to a different DR system. This gives you exactly the outcome in your diagram — two PBR streams to two different DR systems — while staying within a supported configuration.
Requirements to keep in mind:
 
Storage Virtualize 9.1.2 or later (partition split / multi-partition with independent DR)
Each production system supports up to 3 inter-cluster partnerships — your design uses 1 HA (A↔B) + 2 DR (→C, →D) = 3, which is exactly the maximum
One important VMware note if any of these VGs serve an ESXi cluster: a single VMware cluster must stay within one partition — you cannot split hosts of the same ESXi cluster across two partitions. So the VG-to-partition split must align with host/workload boundaries.
 
We have implemented exactly this multi-partition / multi-DR-target design in production, so happy to share the detailed procedure if useful.
Johan van Arendonk's profile image
Johan van Arendonk IBM Champion

Hello Nguyen,

I don't see any partitions drawn in your graphical overview. If all three VGs are located within the same partition, then your setup is not possible.
Policy-based replication links partitions, not volume groups.
 
Perhaps you could create multiple partitions and then distribute the VGs across these two partitions.
Keep in mind, however, that host objects reside within a single partition; if a host requires data from all three VGs, this approach will not work.
 
Have a look at https://www.redbooks.ibm.com/redbooks/pdfs/sg248569.pdf
This book also contains an explanation of 3-site recovery.