Is VMware Clustered VMDK Supported for Qlik Replicate HA on WSFC?

[fahimghauri]

Hi everyone,

I’m planning to implement Qlik Replicate High Availability (HA) using Windows Server Failover Clustering (WSFC). According to Qlik’s documentation for setting up Replicate in a Windows cluster, the shared storage for the Qlik Replicate should be block storage.

VMware provides support for Clustered VMDK disks for WSFC environments, as described in the following VMware document:
<https://www.vmware.com/docs/vmw-vmdk-whitepaper-mmt>

My question is:

➡️ Is a VMware-supported Clustered VMDK (used as shared storage in WSFC) officially supported and compatible with Qlik Replicate’s HA requirements?

Specifically, can the Replicate data directory reside on a clustered VMDK disk shared between the WSFC nodes?

If anyone has implemented this setup—or if Qlik can confirm official support—it would be greatly appreciated.

Thanks in advance!

Fahim

[DesmondWOO]

Hi @fahimghauri ,

I found a case related to VMware Multi‑Writer Shared Disk, and I hope the following information is useful.

==================================

This configuration can be supported under the following conditions:

1. The Replicate service must be configured as a generic WSFC (Windows Server Failover Cluster) service, following the documentation: Qlik Replicate WSFC Setup Guide.
2. All Replicate operations, including command‑line usage, must be performed on the active node only. This is a prerequisite regardless of the storage type.
3. A pre‑script must be added to remove all SQLite .db-shm files. Failure to do so should prevent the service from starting, requiring manual cleanup.

In this scenario, the corruption risk in a WSFC deployment is equivalent to the inherent risk of storage‑level malfunctions that can occur in any environment.

If the above conditions are met, we see no reason to consider this environment unsupported, provided the customer is aware of the slower cluster‑wide locks. The customer should also verify that performance in this environment is satisfactory.
===================================

Regards,
Desmond