Free space elsewhere in a datastore cluster is not sufficient for Storage DRS to move a VM's disks. The destination must satisfy placement constraints, and the VM's automation settings must permit the proposed operation. In vSphere 8.x, intra-VM disk affinity and VM-level Manual or Disabled overrides are two documented reasons for apparently unbalanced storage.
This source-only diagnostic follows KB 432448. No storage policy, recommendation or VM was changed or tested. The numerical example is hypothetical.
Capacity, placement and execution are different questions
Storage DRS works with the member datastores of a datastore cluster. Its placement decision is not equivalent to adding their free bytes and finding any total large enough for the VM. A configured constraint can require several disks to remain together.
The KB identifies the default intra-VM affinity setting, labelled Keep virtual disks together by default, as a potential restriction. The setting expresses a policy for the VM's disks. It should not be confused with a compute DRS VM-host affinity rule, which constrains the hosts on which a VM runs.
After placement is considered, automation introduces another boundary. An existing recommendation and an automatically executed migration are different outcomes. A VM-level Manual or Disabled override may deliberately prevent the behaviour an operator expects from the datastore cluster's general setting.
Inspect the exact VM and datastore cluster
Prerequisites are read access to the datastore-cluster configuration, the affected VM's disk backing and overrides, existing recommendation/task details, and current capacity information. Record vCenter and ESXi versions and the observation time. Do not run a storage migration to discover what configuration applies.
For the affected VM, map each VMDK to its datastore and record its relevant space requirement using the platform's documented capacity fields. Keep provisioned capacity, allocated consumption and free datastore space distinct. A thin disk's maximum size is not automatically its current transfer size or an adequate prediction of future growth.
In Datastore Cluster → Configure → Storage DRS, inspect the existing intra-VM affinity choice. In the datastore cluster's VM Overrides area, inspect the selected VM's automation level. The KB describes these locations in its change procedure; here they are used only to read the current values. Do not save edits.
Then classify what is actually absent: a recommendation, an eligible destination, or automatic execution. Keep the recommendation's own explanation if one exists. A general statement that “DRS did nothing” loses the distinction needed to find the cause.
The KB places the setting at Datastore Cluster → Configure → Storage DRS → Edit → Advanced Options, labelled Default VM affinity (Keep virtual disks together by default). Inspect the current value and cancel without saving.
Worked placement example
Assume a fictional VM has two disks that would require 700 GiB and 500 GiB at the destination for this simplified example. Two candidate datastores each have 800 GiB available. Ignore other overhead only for the arithmetic, and assume no other eligibility problem.
The combined 1,600 GiB appears larger than the 1,200-GiB requirement. However, neither individual datastore can hold both disks. If the effective rule keeps those disks together, the aggregate number does not describe a legal placement. Splitting the disks would satisfy the simplified capacity calculation while violating the stated policy.
Now suppose one datastore has enough space for the entire group and a recommendation exists, but the VM has a Manual override. The explanation has moved from feasibility to execution. More free capacity will not convert that deliberate automation choice into an automatic move.
These are diagnostic models, not predictions of the actual Storage DRS result. Real placement includes other eligibility, capacity and policy considerations. Use the actual recommendation and constraints to confirm which model applies.
Choose the right change owner
If affinity is intentional, the remedy may be additional capacity in a suitable destination or a reviewed placement design. Disabling affinity merely to produce a move changes the VM's storage contract. Establish why the disks were kept together before proposing separation.
If an automation override is intentional, leaving it in place may be correct. A reviewed manual recommendation and a permanent automation-policy change have different scope. Do not replace every override with Fully Automated because a KB lists it as a possible repair.
KB 432448 gives three further concrete checks: independent disks do not support automated live Storage DRS migration; NFS 4.1 I/O load balancing is described as natively unsupported in 8.0; a VM CPU or memory allocation limit of exactly 0 can fail allocation checks. Record those existing values as distinguishing evidence. This does not prescribe changing disk mode, I/O settings or resource limits, and each non-matching branch needs its own investigation.
Acceptance and reversal
Diagnostic acceptance is a VM-to-disk map, the effective affinity and automation values, and an explanation of whether the missing step is legal placement or execution. Later operational acceptance would require the reviewed recommendation to complete and the intended capacity and storage-policy state to remain correct. No such migration was run here.
Reading settings needs no rollback. Restoring a policy value after a future change does not automatically move disks back to their former datastore. Returning placement requires a new supported capacity-checked operation. Do not describe a checkbox reversal as reversal of the storage migration itself.
Learning notes
- Aggregate free space can conceal an impossible affinity-constrained placement.
- A recommendation and permission to execute it are separate facts.
- Storage DRS disk affinity is different from compute DRS host affinity.