FlexVol deep dive

A FlexVol is ONTAP's standard flexible volume: a separately managed WAFL file system inside an aggregate. Its logical size can exceed currently allocated physical space, it can grow or shrink, and NAS clients can reach it through an SVM junction while SAN hosts consume LUN files inside it.

Aggregate containing a FlexVol mounted in an SVM namespace

Volume anatomy

Four coordinates identify a FlexVol: its SVM context, unique volume name within that SVM, containing aggregate, and optional NAS junction path. The aggregate supplies physical blocks; the volume owns files, metadata, Snapshot copies, qtrees and any LUNs.

cluster1::> volume show -vserver svm_nas -volume vol_projects \
  -fields aggregate,size,state,junction-path,space-guarantee
vserver volume       aggregate size state  junction-path space-guarantee
------- ------------ --------- ---- ------ ------------- ---------------
svm_nas vol_projects aggr1     2TB  online /projects     none

A volume can move between aggregates without changing the client namespace; see the volume move runbook.

Space guarantees: volume, file or none

-space-guarantee controls when aggregate space is reserved. volume reserves the volume size from the aggregate; none thin-provisions it. file is a legacy file/LUN-oriented guarantee whose availability and behavior depend on platform and release—check ONTAP 9.19.1 command help before selecting it.

cluster1::> volume create -vserver svm_nas -volume vol_projects \
  -aggregate aggr1 -size 2TB -space-guarantee none \
  -snapshot-reserve 5 -junction-path /projects
[Job 8421] Job succeeded: Successful

cluster1::> volume show -fields size,available,percent-used,space-guarantee
vserver volume       size available percent-used space-guarantee
------- ------------ ---- --------- ------------ ---------------
svm_nas vol_projects 2TB  1.61TB    15%          none
svm_san vol_db       1TB  388GB     60%          volume

Logical size is what the client can address; guaranteed space is what the aggregate commits. A thin volume can therefore report free logical space while the aggregate is running out. Monitor both layers. -fractional-reserve reserves a percentage of a data-protection volume's size for overwrite space; current defaults and valid use cases depend on volume and guarantee type, so check your release rather than blindly forcing 0%.

cluster1::> volume modify -vserver svm_san -volume vol_db \
  -fractional-reserve 100
Volume modify successful on volume vol_db of Vserver svm_san.

Autosize guardrails

grow expands a volume above its grow threshold. grow_shrink also contracts it below the shrink threshold, but shrinking deserves conservative testing because application and Snapshot behavior can make used space oscillate. Set explicit minimum and maximum bounds.

cluster1::> volume modify -vserver svm_nas -volume vol_projects \
  -autosize-mode grow -max-autosize 4TB \
  -autosize-grow-threshold-percent 85
Volume modify successful on volume vol_projects of Vserver svm_nas.

cluster1::> volume autosize -vserver svm_nas -volume vol_projects
Vserver  Volume       Mode Grow% Shrink% Minimum Size Maximum Size
-------- ------------ ---- ----- ------- ------------ ------------
svm_nas  vol_projects grow 85%   50%     2TB          4TB

Some releases expose autosize display/modify through volume autosize as well as volume fields. Use volume modify ? and volume autosize ? on the target release.

Snapshot reserve

-snapshot-reserve sets the percentage reserved for Snapshot copies; it is not a retention schedule. A 5% reserve on a 2 TiB volume separates roughly 102 GiB for Snapshot use, subject to ONTAP accounting. When that reserve fills, Snapshot consumption can spill into active-file-system space according to volume behavior.

cluster1::> volume modify -vserver svm_nas -volume vol_projects -snapshot-reserve 10
Volume modify successful on volume vol_projects of Vserver svm_nas.
cluster1::> volume show -volume vol_projects -fields percent-snapshot-space
vserver volume       percent-snapshot-space
------- ------------ ----------------------
svm_nas vol_projects 10%

Junctions and the SVM namespace

A NAS FlexVol is mounted beneath the SVM root volume. The junction is a namespace pointer, not a directory that copies data. Mounting vol_projects at /projects makes that volume reachable through data LIFs, subject to protocol and export/share configuration.

cluster1::> volume mount -vserver svm_nas -volume vol_projects \
  -junction-path /projects
Volume vol_projects on Vserver svm_nas is mounted at /projects.

cluster1::> volume show -vserver svm_nas -junction
Vserver  Volume        Language Active Junction Junction Path
-------- ------------- -------- ------- -------- -------------
svm_nas  svm_nas_root  C.UTF-8  true    true    /
svm_nas  vol_projects  C.UTF-8  true    true    /projects

For SAN, a LUN is a file-like object allocated inside a FlexVol and presented through an igroup mapping; the volume still controls capacity, Snapshot copies, efficiency and replication. Learn the enclosing SVM model in the SVM guide.

FlexVol versus FlexGroup

Question	FlexVol	FlexGroup
Layout	One volume on one aggregate	One namespace spread across constituent FlexVols
Scale	Bounded by one volume/node placement	Scale-out capacity, throughput and metadata work
Operations	Direct volume management	Manage the FlexGroup; avoid manual constituent changes
Choose when	General NAS/SAN, predictable placement, feature compatibility	Very large NAS namespace or high parallelism

Feature support changes by ONTAP release. Confirm workload requirements against the FlexGroup deep dive.

Quotas, efficiency and clones

Quotas limit or report user, group and qtree consumption inside the volume; they do not reserve aggregate capacity. Dedupe, compression and compaction reduce physical use and affect logical reporting; see storage efficiency. Depending on platform/release, efficiency operates at volume and aggregate layers. A FlexClone starts as a writable, block-sharing clone of a parent Snapshot and consumes new space as blocks diverge.

cluster1::> volume clone create -vserver svm_nas -flexclone vol_projects_test \
  -type RW -parent-volume vol_projects -parent-snapshot nightly.0 \
  -junction-path /projects_test
[Job 8460] Job succeeded: Successful

Offline and delete safely

Confirm shares, exports, LUN mappings, mirrors, clones and application ownership first. Current ONTAP uses volume delete; some older operational language says “destroy,” but volume destroy is not the current command. Deleted volumes may enter the recovery queue unless the chosen force behavior bypasses it.

cluster1::> volume unmount -vserver svm_nas -volume vol_old
cluster1::> volume offline -vserver svm_nas -volume vol_old
Volume "vol_old" on Vserver "svm_nas" is now offline.
cluster1::> volume delete -vserver svm_nas -volume vol_old
Warning: Are you sure you want to delete volume "vol_old"? {y|n}: y
[Job 8498] Job succeeded: Successful

Troubleshooting matrix

Symptom	Prove	Likely fix
Volume full, aggregate free	Volume active/total/Snapshot space and quota report	Resize/autosize, remove approved data or snapshots; do not assume aggregate free space expands it
Aggregate full, thin volume says free	Aggregate available plus all thin commitments	Free/move capacity or add storage; cap autosize contention
Autosize did not grow	Mode, grow threshold, maximum, aggregate headroom	Raise maximum or restore aggregate capacity; inspect EMS/job errors
NAS path absent	Volume state, junction, export/share and data LIF	Mount at the intended junction and validate protocol policy
Unexpected used space	Snapshot, LUN, clone and efficiency accounting	Identify the owner before deleting anything