Operating SystemsStorage Management

Storage Virtualization

LevelAdvanced

Duration90 mins

TopicStorage Management

2 / 5

Volume Groups

The Storage Pool Abstraction

In the previous page, we explored LVM's three-layer architecture and established that Volume Groups (VGs) form the critical middle layer—the bridge between physical storage devices and the logical volumes that applications actually use. Now we dive deep into volume groups: their internal structure, design considerations, operational mechanics, and how they enable the storage flexibility that makes LVM so powerful.

A volume group is fundamentally a storage pool. It aggregates the capacity of one or more physical volumes into a unified resource from which logical volumes can draw space on demand. This pooling concept transforms rigid, device-bound storage into fluid, allocatable capacity.

Consider the profound shift this enables: without volume groups, if you have three 100GB disks, you have three separate 100GB storage spaces. With a volume group encompassing all three, you have a single 300GB pool from which you can carve logical volumes of any size up to 300GB—or create many smaller volumes that collectively use the pooled capacity.

What You Will Learn

By the end of this page, you will understand volume group creation strategies, extent size selection trade-offs, VG metadata structure and redundancy, extending and reducing volume groups, VG naming conventions, clustering considerations for shared storage, and production best practices for volume group design.

Volume Group Fundamentals

A Volume Group serves as the administrative container for LVM storage. It defines the extent size used for allocations, maintains metadata describing all physical and logical volumes, and manages the free space pool from which new logical volumes are created or existing ones are extended.

The Extent Size Decision:

When you create a volume group, you specify an extent size—the fundamental unit of allocation for all storage within that VG. This decision, seemingly minor, has architectural implications that persist for the lifetime of the volume group.

The extent size determines:

Allocation Granularity: You can only allocate storage in whole-extent increments
Maximum Volume Size: Limited by (max LEs per LV) × (extent size)
Metadata Size: More extents mean more mapping entries in metadata
Alignment with Hardware: Optimal performance when aligned with RAID stripes or SSD erase blocks

Extent Size Trade-offs
Extent Size	Max LV Size (64K LEs)	Waste Per LV (max)	Metadata Entries per TB	Best For
1 MB	64 GB	~1 MB	1,048,576	Embedded systems, small deployments
4 MB (default)	256 TB	~4 MB	262,144	General purpose, most servers
8 MB	512 TB	~8 MB	131,072	Large storage systems
16 MB	1 PB	~16 MB	65,536	Enterprise SAN, petabyte scale
64 MB	4 PB	~64 MB	16,384	Massive data lakes, HPC

Why Can't Extent Size Be Changed?

Once a volume group is created, its extent size is immutable. This is because:

Physical extents are pre-allocated: The entire PV is divided into PEs at initialization
Logical extent mappings assume fixed size: LE-to-PE mappings would become invalid
Segment calculations depend on extent size: All segment start/length values are in extents
Device-mapper tables are derived from extent mappings: The kernel-level mapping would break

If you need a different extent size, you must create a new volume group and migrate data to it—a process that, while possible with pvmove, requires careful planning and potentially significant time.

Extent Size Recommendation

For modern systems, the 4MB default is appropriate for most use cases. Only increase extent size if you're building storage systems at petabyte scale or have specific alignment requirements. Decreasing below 4MB is rarely beneficial and increases metadata overhead substantially.

Volume Group Metadata Structure

Volume group metadata is the authoritative record of the entire LVM configuration. It contains definitions for all physical volumes in the VG, all logical volumes, their segment mappings, and historical information. Understanding metadata structure is essential for troubleshooting and disaster recovery.

Metadata Storage Locations:

By default, LVM stores a complete copy of VG metadata on every physical volume in the volume group. This redundancy ensures that:

Loss of any single PV doesn't result in metadata loss
Any surviving PV can be used to reconstruct the VG configuration
The VG can operate even with degraded PV availability (for unmapped LVs)

The metadata area (MDA) is typically located at the beginning of the PV, after the label sector. The default MDA size is approximately 1MB, but this can be adjusted during pvcreate.

VG Metadata Analysis
Bash
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
#!/bin/bash
# Examining Volume Group Metadata
 
# Display raw metadata from a VG (extremely useful for debugging)
vgcfgbackup -f /tmp/vg_storage_backup.txt vg_storage
 
# View the metadata (it's in a human-readable ASCII format)
cat /tmp/vg_storage_backup.txt
 
# Example output structure:
# vg_storage {
#     id = "aB1CdE-2fGh-3iJk-4LmN-5oPq-6RsT-7UvWx"
#     seqno = 42
#     format = "lvm2"
#     status = ["RESIZEABLE", "READ", "WRITE"]
#     flags = []
#     extent_size = 8192          # In sectors (4MB = 8192 * 512)
#     max_lv = 0
#     max_pv = 0
#     metadata_copies = 0
#
#     physical_volumes { ... }
#     logical_volumes { ... }
# }
 
# Check metadata area size on a PV
pvs -o+pv_mda_size,pv_mda_free /dev/sdb
# Output:
#   PV         VG         PSize   PFree  PMdaSize  PMdaFree
#   /dev/sdb   vg_storage 100.00g 20.00g  1020.00k   508.00k
 
# Display where metadata is stored
pvck --dump metadata /dev/sdb
# Shows detailed metadata area information
 
# === METADATA AREA CONFIGURATION ===
 
# Create PV with larger metadata area (2 copies, 2MB each)
pvcreate --metadatasize 2m --pvmetadatacopies 2 /dev/sdd
 
# Create PV with NO metadata area (uses another PV's metadata)
# Useful for data-only PVs in large configurations
pvcreate --metadatacopies 0 /dev/sdd
 
# === METADATA OPERATIONS ===
 
# Check metadata consistency
vgck vg_storage
 
# Display metadata sequence number (increments with each change)
vgs -o vg_name,vg_seqno vg_storage
 
# Restore metadata from backup (disaster recovery)
vgcfgrestore -f /tmp/vg_storage_backup.txt vg_storage
 
# Force metadata refresh from disk
pvscan --cache
 
# View metadata history (if archived)
ls /etc/lvm/archive/
# Files named: vg_storage_00042-xxxxxxxx.vg

Metadata Sequence Numbers:

Every time VG metadata is modified (creating an LV, extending a volume, adding a PV), the sequence number increments. This versioning enables:

Consistency Checking: LVM can detect if PVs have divergent metadata versions
Recovery: The most recent sequence number identifies the authoritative metadata
Archiving: /etc/lvm/archive/ stores sequential metadata backups
Debugging: You can trace the history of configuration changes

Metadata Formats:

LVM2 uses a text-based metadata format that is:

Human-readable: You can view and understand the configuration
Portable: The same format works across all Linux versions
Versionless: Forward-compatible by ignoring unknown fields
Compact: Uses segment representation to minimize size

Metadata Backup Is Critical

Always run 'vgcfgbackup' before major storage operations and store the backup off-system. LVM automatically archives metadata to /etc/lvm/archive/, but this won't help if the root filesystem is on the affected VG. Having an external metadata backup can save hours of recovery effort.

Creating Volume Groups

Volume group creation involves several strategic decisions that affect long-term manageability. While the basic vgcreate command is straightforward, understanding all options enables optimal configuration for specific use cases.

Basic VG Creation:

The simplest form creates a VG with default settings:

Volume Group Creation
Bash
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
#!/bin/bash
# Volume Group Creation Examples
 
# === BASIC CREATION ===
 
# Create VG with single PV (defaults to 4MB extent size)
vgcreate vg_data /dev/sdb
 
# Create VG with multiple PVs
vgcreate vg_data /dev/sdb /dev/sdc /dev/sdd
 
# === EXTENT SIZE CONFIGURATION ===
 
# Create VG with 8MB extents (larger systems)
vgcreate -s 8M vg_large /dev/sdb /dev/sdc
 
# Create VG with 16MB extents (enterprise scale)
vgcreate -s 16M vg_enterprise /dev/sd[b-h]
 
# Create VG with 1MB extents (fine-grained control, small systems)
vgcreate -s 1M vg_small /dev/sdb
 
# === ALLOCATION POLICY ===
 
# Create VG with contiguous allocation policy (for performance)
vgcreate --alloc contiguous vg_perf /dev/nvme0n1p2
 
# Allocation policies:
# - inherit: Use default policy
# - contiguous: Require physically contiguous extents
# - cling: Prefer same PV as existing allocations
# - normal: Default, balance between policies
# - anywhere: Allow any available extent
 
# === METADATA CONFIGURATION ===
 
# Create VG with maximum metadata copies (highest redundancy)
vgcreate --vgmetadatacopies all vg_critical /dev/sd[b-e]
 
# Create VG with single metadata copy (minimum redundancy)
vgcreate --vgmetadatacopies 1 vg_temp /dev/sdb
 
# Set specific number of metadata copies
vgcreate --vgmetadatacopies 3 vg_balanced /dev/sd[b-f]
 
# === CLUSTERING (for shared storage) ===
 
# Create clustered VG (requires clvmd or lvmlockd)
vgcreate --shared vg_cluster /dev/mapper/shared_lun
 
# Create VG with system ID for exclusive access
vgcreate --systemid "server1" vg_server1 /dev/sdb
 
# === ADVANCED OPTIONS ===
 
# Auto-activate VG at boot
vgcreate --setautoactivation y vg_boot /dev/sdb
 
# Set maximum number of LVs (0 = unlimited, default)
vgcreate --maxlogicalvolumes 100 vg_limited /dev/sdb
 
# Set maximum number of PVs (0 = unlimited, default)
vgcreate --maxphysicalvolumes 10 vg_bounded /dev/sdb
 
# === VERIFY CREATION ===
 
# Display detailed information about new VG
vgdisplay vg_data
 
# Show VG with all available columns
vgs -o+vg_extent_size,vg_extent_count,vg_free_count vg_data
 
# Verify metadata consistency
vgck vg_data

VG Naming Conventions:

Choosing clear, consistent VG names improves system administration. Common conventions include:

Pattern	Example	Use Case
Purpose-based	`vg_database`, `vg_logs`	Dedicated VGs for specific workloads
Host-based	`vg_<hostname>`	When VGs may move between systems
Storage-tier	`vg_ssd`, `vg_hdd`, `vg_nvme`	Differentiating performance tiers
Datacenter	`vg_dc1_rack3`	Large-scale deployments
Combined	`vg_prod_db_ssd`	Complex environments

Avoid names that might conflict with device nodes or contain special characters. Use lowercase with underscores for maximum compatibility.

VG Creation Prerequisites

Before creating a volume group, ensure all target devices are initialized as physical volumes with 'pvcreate'. If a device contains existing partitions or file systems, pvcreate will warn and may require force (-f) flag. Always verify you're operating on the intended devices before proceeding.

Extending and Reducing Volume Groups

One of LVM's most powerful capabilities is dynamic volume group modification. You can add new physical volumes to increase capacity or remove physical volumes during hardware decommissioning—often without any service interruption.

Extending Volume Groups:

Adding capacity to a volume group is a straightforward, non-disruptive operation:

VG Extension Operations
Bash
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
#!/bin/bash
# Extending Volume Groups
 
# === ADDING NEW PHYSICAL VOLUMES ===
 
# Step 1: Initialize the new device as a PV
pvcreate /dev/sdd
 
# Step 2: Add PV to existing VG
vgextend vg_data /dev/sdd
 
# Verify the extension
vgdisplay vg_data | grep -E 'VG Size|Free PE'
# Before: VG Size 200.00 GiB, Free PE / Size 0 / 0
# After:  VG Size 300.00 GiB, Free PE / Size 25600 / 100.00 GiB
 
# One-liner: Combine pvcreate and vgextend
pvcreate /dev/sde && vgextend vg_data /dev/sde
 
# Add multiple PVs at once
vgextend vg_data /dev/sdf /dev/sdg /dev/sdh
 
# === ADDING PV WITH SPECIFIC MDA OPTIONS ===
 
# Add PV without metadata (data-only, relies on other PVs for metadata)
pvcreate --metadatacopies 0 /dev/sdi
vgextend vg_data /dev/sdi
 
# Add PV with extra metadata copies
pvcreate --metadatasize 2M --pvmetadatacopies 2 /dev/sdj
vgextend vg_data /dev/sdj
 
# === POST-EXTENSION OPERATIONS ===
 
# Now you can extend logical volumes using the new space
lvextend -L +50G -r /dev/vg_data/lv_database
 
# Or create new logical volumes
lvcreate -L 100G -n lv_archive vg_data /dev/sdd
 
# === CAPACITY PLANNING ===
 
# Check how much space was added from new PV
pvs /dev/sdd -o pv_name,pv_size,pv_used,pv_free
 
# Show overall VG utilization
vgs vg_data -o vg_name,vg_size,vg_free,pv_count

Reducing Volume Groups:

Removing physical volumes from a volume group requires more care. You must first ensure no data resides on the PV being removed, either by migrating data or because the PV was never used.

VG Reduction Operations
Bash
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
#!/bin/bash
# Reducing Volume Groups
 
# === SAFE REDUCTION (PV has no allocated extents) ===
 
# Check if PV has allocated extents
pvs -o pv_name,pv_size,pv_used,pv_pe_count,pv_pe_alloc_count /dev/sdd
 
# If Alloc PE is 0, safe to remove
vgreduce vg_data /dev/sdd
 
# Wipe LVM label from removed PV
pvremove /dev/sdd
 
# === REDUCTION WITH DATA MIGRATION ===
 
# Step 1: Move all extents from sdd to other PVs in VG
# This is an online operation but may take significant time
pvmove /dev/sdd
 
# The command will show progress:
#   /dev/sdd: Moved: 12.5%
#   /dev/sdd: Moved: 25.0%
#   ...
#   /dev/sdd: Moved: 100.0%
 
# Step 2: Verify PV is now empty
pvs /dev/sdd -o pv_used
# Should show 0
 
# Step 3: Remove from VG
vgreduce vg_data /dev/sdd
 
# Step 4: Remove PV label
pvremove /dev/sdd
 
# === TARGETED DATA MIGRATION ===
 
# Move only specific LV's extents off the PV
pvmove -n lv_database /dev/sdd /dev/sde
 
# Move data to specific destination PVs only
pvmove /dev/sdd /dev/sde /dev/sdf
 
# === HANDLING ERRORS ===
 
# If VG has missing PV (disk failed), reduce with force
vgreduce --removemissing vg_data
 
# If PV is inaccessible, force reduction
vgreduce --force vg_data /dev/sdd
 
# === PRE-REDUCTION VALIDATION ===
 
# List all LV segments showing which PV they use
lvs --segments -o lv_name,seg_start,seg_size,devices
 
# Find all LVs with extents on specific PV
pvs -o pv_name,lv_name,seg_all /dev/sdd

Data Migration Time

The pvmove operation physically copies data between devices. For a fully-utilized 1TB disk, this can take hours depending on I/O bandwidth. Plan pvmove operations during maintenance windows, even though they're online-safe. Use 'pvmove -b' for background operation and monitor with 'lvs -a -o +devices %vg_name'.

Volume Group Allocation Strategies

When logical volumes request space, LVM must decide which physical extents to allocate. The allocation policy controls this selection, affecting both performance and flexibility. Understanding allocation behavior is crucial for optimal storage design.

Allocation Policies in Depth:

LVM Allocation Policies

•contiguous: All physical extents must be adjacent on a single PV. Maximizes sequential I/O performance but severely limits flexibility. Fails if contiguous space isn't available, even if total free space suffices.
•cling: Tries to place new extents on the same PV as existing extents of the LV. Falls back to other PVs if necessary. Useful for reducing inter-PV seeks.
•cling_by_tags: Like cling, but uses PV tags to group 'similar' PVs. Enables cling behavior across multiple PVs with the same tag (e.g., same storage tier).
•normal: Default policy. Prefers to use fewer PVs and avoids splitting across many devices. Balances between contiguity and flexibility.
•anywhere: Use any available extent without preference. Enables maximum space utilization but may cause fragmentation. Required for striped volumes spanning multiple PVs.

Converting Mermaid diagram...

Performance Implications:

Policy	Sequential I/O	Random I/O	Space Efficiency	Use Case
contiguous	Excellent	Excellent	Poor	High-performance databases, sequential workloads
cling	Very Good	Good	Good	General database, consistent performance
normal	Good	Good	Very Good	Default, most workloads
anywhere	Variable	Fair	Excellent	Space-constrained environments

Tagging for Advanced Allocation:

PV tags enable sophisticated allocation control by grouping physical volumes with similar characteristics:

PV Tagging for Allocation Control
Bash
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
#!/bin/bash
# Using PV Tags for Allocation Control
 
# Tag PVs by storage tier
pvchange --addtag @tier_ssd /dev/nvme0n1p2 /dev/nvme1n1p2
pvchange --addtag @tier_hdd /dev/sdb /dev/sdc /dev/sdd
 
# Tag PVs by physical location (for fault domains)
pvchange --addtag @rack_a /dev/nvme0n1p2 /dev/sdb
pvchange --addtag @rack_b /dev/nvme1n1p2 /dev/sdc
 
# View tags on PVs
pvs -o pv_name,pv_tags
 
# Create LV only on SSD-tier PVs
lvcreate -L 50G -n lv_fast vg_mixed @tier_ssd
 
# Create LV only on HDD-tier PVs
lvcreate -L 500G -n lv_archive vg_mixed @tier_hdd
 
# Extend LV using only same-tier storage (cling_by_tags)
lvextend --alloc cling_by_tags -L +25G /dev/vg_mixed/lv_fast
 
# Create mirrored LV with legs on different racks (fault isolation)
lvcreate --type raid1 -L 100G -n lv_critical --alloc anywhere \
    vg_mixed @rack_a @rack_b
 
# Remove a tag
pvchange --deltag @tier_ssd /dev/nvme0n1p2
 
# List all available tags in system
vgs -o tags
pvs -o tags

Strategic Tagging

Use tags to implement storage tiering without separate volume groups. A single VG containing both SSDs (tagged @ssd) and HDDs (tagged @hdd) allows you to place high-performance LVs on SSDs and archival LVs on HDDs, while still enabling pvmove between tiers as requirements change.

Volume Group Operations

Beyond creation and extension, volume groups support various administrative operations for maintenance, migration, and troubleshooting.

Activation and Deactivation:

Volume groups and their logical volumes must be activated before use. Activation loads the device-mapper mappings into the kernel, making the LVs accessible as block devices.

VG Administrative Operations
Bash
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
#!/bin/bash
# Volume Group Administrative Operations
 
# === ACTIVATION CONTROL ===
 
# Activate all LVs in a VG
vgchange -ay vg_data
 
# Deactivate all LVs in a VG (must unmount first!)
vgchange -an vg_data
 
# Activate specific LV only
lvchange -ay /dev/vg_data/lv_database
 
# Check activation status
lvs -o lv_name,lv_attr vg_data
# Attribute field 5: 'a' = active, '-' = inactive
 
# Set auto-activation on boot
vgchange --setautoactivation y vg_data
 
# Disable auto-activation (for manual control)
vgchange --setautoactivation n vg_data
 
# === VG RENAME ===
 
# Rename volume group (all LVs automatically get new paths)
vgrename vg_data vg_storage
 
# Verify new name
vgs vg_storage
 
# Update fstab entries for new device paths!
 
# === VG EXPORT/IMPORT (for moving between systems) ===
 
# Step 1: Deactivate VG
vgchange -an vg_portable
 
# Step 2: Export VG (marks it as foreign)
vgexport vg_portable
 
# === On new system ===
 
# Step 3: Scan for all PVs
pvscan
 
# Step 4: Import the VG
vgimport vg_portable
 
# Step 5: Activate
vgchange -ay vg_portable
 
# === VG SPLIT (separate PVs into new VG) ===
 
# Move LV and its PVs to new VG (must specify all necessary PVs)
vgsplit vg_data vg_archive /dev/sdd /dev/sde -n lv_old_data
 
# Note: Source VG must have no LVs with extents on moved PVs
# except for the ones being moved
 
# === VG MERGE (combine VGs) ===
 
# Merge vg_archive into vg_data
# Deactivate source VG first
vgchange -an vg_archive
vgmerge vg_data vg_archive
 
# === VG ATTRIBUTE CHANGES ===
 
# Make VG read-only (all LVs become read-only)
vgchange -p r vg_archive
 
# Restore read-write
vgchange -p rw vg_archive
 
# Set VG maximum LV limit
vgchange -l 50 vg_data
 
# Remove the limit (set to 0 = unlimited)
vgchange -l 0 vg_data
 
# === RECOVERY OPERATIONS ===
 
# Repair VG with missing PVs
vgreduce --removemissing --force vg_data
 
# Restore VG from metadata backup
vgcfgrestore -f /backup/vg_data.txt vg_data
 
# Rebuild metadata from disk scan
vgscan --mknodes
 
# Force reread of all LVM metadata
pvscan --cache

VG Export/Import Workflow:

When moving storage between systems (e.g., moving a disk array to a new server), the export/import workflow ensures clean handoff:

Deactivate: Ensures no I/O is in progress
Export: Marks metadata indicating VG is in transit
Physical move: Relocate the disks/LUNs
Scan: New system discovers the PVs
Import: Clears the export flag, registers VG on new system
Activate: Makes LVs accessible

This workflow prevents accidentally accessing the same storage from multiple systems, which could cause data corruption.

VG Merge Prerequisites

Two volume groups can only be merged if they have identical extent sizes. If extent sizes differ, you must create a new VG with the desired size and migrate data. Also ensure no LV name conflicts exist between the VGs before merging.

Production Best Practices

Designing volume group configurations for production environments requires balancing flexibility, performance, reliability, and manageability. The following practices derive from years of enterprise deployment experience.

Design Recommendations

•Separate VGs by failure domain: Don't mix local disks and SAN storage in one VG; failures affect differently
•Separate VGs by performance tier: Keep SSDs and HDDs in separate VGs for predictable performance
•Use descriptive VG names: Names like 'vg_prod_db_ssd' are self-documenting
•Reserve 15-20% free space: Enables snapshots, emergencies, and pvmove operations
•Prefer fewer, larger VGs: Reduces administrative overhead; use tags for allocation control
•Document extent size rationale: Future administrators need to understand why 8MB vs 4MB was chosen

Common Mistakes to Avoid

•Allocating 100% of VG space: Leaves no room for snapshots or growth
•Mixing boot and data VGs: Boot issues shouldn't affect data availability
•Using random extent sizes: Stick to defaults unless specific requirements exist
•Ignoring metadata redundancy: At least 2 PVs should have metadata copies
•No external metadata backup: Losing all PVs means losing everything
•Complex VG topologies for small deployments: Adds complexity without benefit

Monitoring and Capacity Planning:

Proactive monitoring prevents emergency situations:

VG Monitoring Script
Bash
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
#!/bin/bash
# LVM Monitoring Script - Add to cron for regular checks
 
LOG="/var/log/lvm_monitor.log"
THRESHOLD=85  # Alert when VG is 85% full
 
echo "=== LVM Health Check: $(date) ===" >> $LOG
 
# Check VG capacity
for vg in $(vgs --noheadings -o vg_name); do
    total=$(vgs --noheadings --units g -o vg_size $vg | tr -d ' g')
    free=$(vgs --noheadings --units g -o vg_free $vg | tr -d ' g')
    used_pct=$(echo "scale=0; (($total - $free) / $total) * 100" | bc)
    
    if [ $used_pct -ge $THRESHOLD ]; then
        echo "WARNING: VG $vg is ${used_pct}% full" >> $LOG
        # Send alert (email, Slack, PagerDuty, etc.)
        # mail -s "LVM Alert: $vg at ${used_pct}%" admin@example.com
    fi
done
 
# Check for missing PVs
pvscan 2>&1 | grep -i "missing" >> $LOG
 
# Check metadata sequence consistency
for vg in $(vgs --noheadings -o vg_name); do
    vgck $vg 2>> $LOG
done
 
# List PVs with low free space (individual PV capacity)
pvs --noheadings -o pv_name,pv_size,pv_free | while read pv size free; do
    # Alert if individual PV is nearly full (affects allocation locality)
    # This is informational for cling policy effectiveness
    echo "PV: $pv  Size: $size  Free: $free" >> $LOG
done

Documentation Habit

Maintain a storage layout document that includes: VG names and purposes, extent sizes with rationale, which LVs reside in each VG, expected growth patterns, and metadata backup locations. This document is invaluable during incidents and when onboarding new team members.

Summary: Volume Groups

Volume Groups form the aggregation layer of LVM, pooling physical storage into flexible capacity pools. Let's consolidate the essential concepts covered:

Key Takeaways

•Extent size is permanent: Choose carefully at VG creation; 4MB default suits most workloads
•Metadata redundancy is automatic: Every PV stores VG metadata by default, providing resilience
•VGs can grow dynamically: Adding PVs with vgextend immediately makes capacity available
•PV removal requires data migration: Use pvmove before vgreduce to safely remove hardware
•Allocation policies control placement: From contiguous (performance) to anywhere (flexibility)
•Tags enable sophisticated control: Label PVs by tier, location, or purpose for targeted allocation
•Export/Import enables portability: Move VGs between systems cleanly and safely
•Reserve free space: 15-20% unallocated enables snapshots and emergency capacity

What's Next:

With volume groups understood, we now ascend to the layer that applications actually interact with: Logical Volumes. The next page explores LV creation, resizing, types (linear, striped, mirrored), and the pivotal relationship between logical volumes and the file systems that inhabit them.

Page Complete

You now possess comprehensive knowledge of LVM Volume Groups—from creation through extension, reduction, and production deployment. This understanding prepares you for exploring Logical Volumes, where the storage abstraction meets the file system layer.

2 / 5

Loading learning content...

Operating SystemsStorage Management

Storage Virtualization

LevelAdvanced

Duration90 mins

TopicStorage Management

2 / 5

Volume Groups

The Storage Pool Abstraction

What You Will Learn

Volume Group Fundamentals

The Extent Size Decision:

The extent size determines:

Allocation Granularity: You can only allocate storage in whole-extent increments
Maximum Volume Size: Limited by (max LEs per LV) × (extent size)
Metadata Size: More extents mean more mapping entries in metadata
Alignment with Hardware: Optimal performance when aligned with RAID stripes or SSD erase blocks

Extent Size Trade-offs
Extent Size	Max LV Size (64K LEs)	Waste Per LV (max)	Metadata Entries per TB	Best For
1 MB	64 GB	~1 MB	1,048,576	Embedded systems, small deployments
4 MB (default)	256 TB	~4 MB	262,144	General purpose, most servers
8 MB	512 TB	~8 MB	131,072	Large storage systems
16 MB	1 PB	~16 MB	65,536	Enterprise SAN, petabyte scale
64 MB	4 PB	~64 MB	16,384	Massive data lakes, HPC

Why Can't Extent Size Be Changed?

Once a volume group is created, its extent size is immutable. This is because:

Physical extents are pre-allocated: The entire PV is divided into PEs at initialization
Logical extent mappings assume fixed size: LE-to-PE mappings would become invalid
Segment calculations depend on extent size: All segment start/length values are in extents
Device-mapper tables are derived from extent mappings: The kernel-level mapping would break

Extent Size Recommendation

Volume Group Metadata Structure

Metadata Storage Locations:

By default, LVM stores a complete copy of VG metadata on every physical volume in the volume group. This redundancy ensures that:

Loss of any single PV doesn't result in metadata loss
Any surviving PV can be used to reconstruct the VG configuration
The VG can operate even with degraded PV availability (for unmapped LVs)

The metadata area (MDA) is typically located at the beginning of the PV, after the label sector. The default MDA size is approximately 1MB, but this can be adjusted during pvcreate.

VG Metadata Analysis
Bash
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
#!/bin/bash
# Examining Volume Group Metadata
 
# Display raw metadata from a VG (extremely useful for debugging)
vgcfgbackup -f /tmp/vg_storage_backup.txt vg_storage
 
# View the metadata (it's in a human-readable ASCII format)
cat /tmp/vg_storage_backup.txt
 
# Example output structure:
# vg_storage {
#     id = "aB1CdE-2fGh-3iJk-4LmN-5oPq-6RsT-7UvWx"
#     seqno = 42
#     format = "lvm2"
#     status = ["RESIZEABLE", "READ", "WRITE"]
#     flags = []
#     extent_size = 8192          # In sectors (4MB = 8192 * 512)
#     max_lv = 0
#     max_pv = 0
#     metadata_copies = 0
#
#     physical_volumes { ... }
#     logical_volumes { ... }
# }
 
# Check metadata area size on a PV
pvs -o+pv_mda_size,pv_mda_free /dev/sdb
# Output:
#   PV         VG         PSize   PFree  PMdaSize  PMdaFree
#   /dev/sdb   vg_storage 100.00g 20.00g  1020.00k   508.00k
 
# Display where metadata is stored
pvck --dump metadata /dev/sdb
# Shows detailed metadata area information
 
# === METADATA AREA CONFIGURATION ===
 
# Create PV with larger metadata area (2 copies, 2MB each)
pvcreate --metadatasize 2m --pvmetadatacopies 2 /dev/sdd
 
# Create PV with NO metadata area (uses another PV's metadata)
# Useful for data-only PVs in large configurations
pvcreate --metadatacopies 0 /dev/sdd
 
# === METADATA OPERATIONS ===
 
# Check metadata consistency
vgck vg_storage
 
# Display metadata sequence number (increments with each change)
vgs -o vg_name,vg_seqno vg_storage
 
# Restore metadata from backup (disaster recovery)
vgcfgrestore -f /tmp/vg_storage_backup.txt vg_storage
 
# Force metadata refresh from disk
pvscan --cache
 
# View metadata history (if archived)
ls /etc/lvm/archive/
# Files named: vg_storage_00042-xxxxxxxx.vg

Metadata Sequence Numbers:

Every time VG metadata is modified (creating an LV, extending a volume, adding a PV), the sequence number increments. This versioning enables:

Consistency Checking: LVM can detect if PVs have divergent metadata versions
Recovery: The most recent sequence number identifies the authoritative metadata
Archiving: /etc/lvm/archive/ stores sequential metadata backups
Debugging: You can trace the history of configuration changes

Metadata Formats:

LVM2 uses a text-based metadata format that is:

Human-readable: You can view and understand the configuration
Portable: The same format works across all Linux versions
Versionless: Forward-compatible by ignoring unknown fields
Compact: Uses segment representation to minimize size

Metadata Backup Is Critical

Creating Volume Groups

Basic VG Creation:

The simplest form creates a VG with default settings:

Volume Group Creation
Bash
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
#!/bin/bash
# Volume Group Creation Examples
 
# === BASIC CREATION ===
 
# Create VG with single PV (defaults to 4MB extent size)
vgcreate vg_data /dev/sdb
 
# Create VG with multiple PVs
vgcreate vg_data /dev/sdb /dev/sdc /dev/sdd
 
# === EXTENT SIZE CONFIGURATION ===
 
# Create VG with 8MB extents (larger systems)
vgcreate -s 8M vg_large /dev/sdb /dev/sdc
 
# Create VG with 16MB extents (enterprise scale)
vgcreate -s 16M vg_enterprise /dev/sd[b-h]
 
# Create VG with 1MB extents (fine-grained control, small systems)
vgcreate -s 1M vg_small /dev/sdb
 
# === ALLOCATION POLICY ===
 
# Create VG with contiguous allocation policy (for performance)
vgcreate --alloc contiguous vg_perf /dev/nvme0n1p2
 
# Allocation policies:
# - inherit: Use default policy
# - contiguous: Require physically contiguous extents
# - cling: Prefer same PV as existing allocations
# - normal: Default, balance between policies
# - anywhere: Allow any available extent
 
# === METADATA CONFIGURATION ===
 
# Create VG with maximum metadata copies (highest redundancy)
vgcreate --vgmetadatacopies all vg_critical /dev/sd[b-e]
 
# Create VG with single metadata copy (minimum redundancy)
vgcreate --vgmetadatacopies 1 vg_temp /dev/sdb
 
# Set specific number of metadata copies
vgcreate --vgmetadatacopies 3 vg_balanced /dev/sd[b-f]
 
# === CLUSTERING (for shared storage) ===
 
# Create clustered VG (requires clvmd or lvmlockd)
vgcreate --shared vg_cluster /dev/mapper/shared_lun
 
# Create VG with system ID for exclusive access
vgcreate --systemid "server1" vg_server1 /dev/sdb
 
# === ADVANCED OPTIONS ===
 
# Auto-activate VG at boot
vgcreate --setautoactivation y vg_boot /dev/sdb
 
# Set maximum number of LVs (0 = unlimited, default)
vgcreate --maxlogicalvolumes 100 vg_limited /dev/sdb
 
# Set maximum number of PVs (0 = unlimited, default)
vgcreate --maxphysicalvolumes 10 vg_bounded /dev/sdb
 
# === VERIFY CREATION ===
 
# Display detailed information about new VG
vgdisplay vg_data
 
# Show VG with all available columns
vgs -o+vg_extent_size,vg_extent_count,vg_free_count vg_data
 
# Verify metadata consistency
vgck vg_data

VG Naming Conventions:

Choosing clear, consistent VG names improves system administration. Common conventions include:

Pattern	Example	Use Case
Purpose-based	`vg_database`, `vg_logs`	Dedicated VGs for specific workloads
Host-based	`vg_<hostname>`	When VGs may move between systems
Storage-tier	`vg_ssd`, `vg_hdd`, `vg_nvme`	Differentiating performance tiers
Datacenter	`vg_dc1_rack3`	Large-scale deployments
Combined	`vg_prod_db_ssd`	Complex environments

Avoid names that might conflict with device nodes or contain special characters. Use lowercase with underscores for maximum compatibility.

VG Creation Prerequisites

Extending and Reducing Volume Groups

Extending Volume Groups:

Adding capacity to a volume group is a straightforward, non-disruptive operation:

VG Extension Operations
Bash
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
#!/bin/bash
# Extending Volume Groups
 
# === ADDING NEW PHYSICAL VOLUMES ===
 
# Step 1: Initialize the new device as a PV
pvcreate /dev/sdd
 
# Step 2: Add PV to existing VG
vgextend vg_data /dev/sdd
 
# Verify the extension
vgdisplay vg_data | grep -E 'VG Size|Free PE'
# Before: VG Size 200.00 GiB, Free PE / Size 0 / 0
# After:  VG Size 300.00 GiB, Free PE / Size 25600 / 100.00 GiB
 
# One-liner: Combine pvcreate and vgextend
pvcreate /dev/sde && vgextend vg_data /dev/sde
 
# Add multiple PVs at once
vgextend vg_data /dev/sdf /dev/sdg /dev/sdh
 
# === ADDING PV WITH SPECIFIC MDA OPTIONS ===
 
# Add PV without metadata (data-only, relies on other PVs for metadata)
pvcreate --metadatacopies 0 /dev/sdi
vgextend vg_data /dev/sdi
 
# Add PV with extra metadata copies
pvcreate --metadatasize 2M --pvmetadatacopies 2 /dev/sdj
vgextend vg_data /dev/sdj
 
# === POST-EXTENSION OPERATIONS ===
 
# Now you can extend logical volumes using the new space
lvextend -L +50G -r /dev/vg_data/lv_database
 
# Or create new logical volumes
lvcreate -L 100G -n lv_archive vg_data /dev/sdd
 
# === CAPACITY PLANNING ===
 
# Check how much space was added from new PV
pvs /dev/sdd -o pv_name,pv_size,pv_used,pv_free
 
# Show overall VG utilization
vgs vg_data -o vg_name,vg_size,vg_free,pv_count

Reducing Volume Groups:

Removing physical volumes from a volume group requires more care. You must first ensure no data resides on the PV being removed, either by migrating data or because the PV was never used.

VG Reduction Operations
Bash
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
#!/bin/bash
# Reducing Volume Groups
 
# === SAFE REDUCTION (PV has no allocated extents) ===
 
# Check if PV has allocated extents
pvs -o pv_name,pv_size,pv_used,pv_pe_count,pv_pe_alloc_count /dev/sdd
 
# If Alloc PE is 0, safe to remove
vgreduce vg_data /dev/sdd
 
# Wipe LVM label from removed PV
pvremove /dev/sdd
 
# === REDUCTION WITH DATA MIGRATION ===
 
# Step 1: Move all extents from sdd to other PVs in VG
# This is an online operation but may take significant time
pvmove /dev/sdd
 
# The command will show progress:
#   /dev/sdd: Moved: 12.5%
#   /dev/sdd: Moved: 25.0%
#   ...
#   /dev/sdd: Moved: 100.0%
 
# Step 2: Verify PV is now empty
pvs /dev/sdd -o pv_used
# Should show 0
 
# Step 3: Remove from VG
vgreduce vg_data /dev/sdd
 
# Step 4: Remove PV label
pvremove /dev/sdd
 
# === TARGETED DATA MIGRATION ===
 
# Move only specific LV's extents off the PV
pvmove -n lv_database /dev/sdd /dev/sde
 
# Move data to specific destination PVs only
pvmove /dev/sdd /dev/sde /dev/sdf
 
# === HANDLING ERRORS ===
 
# If VG has missing PV (disk failed), reduce with force
vgreduce --removemissing vg_data
 
# If PV is inaccessible, force reduction
vgreduce --force vg_data /dev/sdd
 
# === PRE-REDUCTION VALIDATION ===
 
# List all LV segments showing which PV they use
lvs --segments -o lv_name,seg_start,seg_size,devices
 
# Find all LVs with extents on specific PV
pvs -o pv_name,lv_name,seg_all /dev/sdd

Data Migration Time

Volume Group Allocation Strategies

Allocation Policies in Depth:

LVM Allocation Policies

•contiguous: All physical extents must be adjacent on a single PV. Maximizes sequential I/O performance but severely limits flexibility. Fails if contiguous space isn't available, even if total free space suffices.
•cling: Tries to place new extents on the same PV as existing extents of the LV. Falls back to other PVs if necessary. Useful for reducing inter-PV seeks.
•cling_by_tags: Like cling, but uses PV tags to group 'similar' PVs. Enables cling behavior across multiple PVs with the same tag (e.g., same storage tier).
•normal: Default policy. Prefers to use fewer PVs and avoids splitting across many devices. Balances between contiguity and flexibility.
•anywhere: Use any available extent without preference. Enables maximum space utilization but may cause fragmentation. Required for striped volumes spanning multiple PVs.

Converting Mermaid diagram...

Performance Implications:

Policy	Sequential I/O	Random I/O	Space Efficiency	Use Case
contiguous	Excellent	Excellent	Poor	High-performance databases, sequential workloads
cling	Very Good	Good	Good	General database, consistent performance
normal	Good	Good	Very Good	Default, most workloads
anywhere	Variable	Fair	Excellent	Space-constrained environments

Tagging for Advanced Allocation:

PV tags enable sophisticated allocation control by grouping physical volumes with similar characteristics:

PV Tagging for Allocation Control
Bash
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
#!/bin/bash
# Using PV Tags for Allocation Control
 
# Tag PVs by storage tier
pvchange --addtag @tier_ssd /dev/nvme0n1p2 /dev/nvme1n1p2
pvchange --addtag @tier_hdd /dev/sdb /dev/sdc /dev/sdd
 
# Tag PVs by physical location (for fault domains)
pvchange --addtag @rack_a /dev/nvme0n1p2 /dev/sdb
pvchange --addtag @rack_b /dev/nvme1n1p2 /dev/sdc
 
# View tags on PVs
pvs -o pv_name,pv_tags
 
# Create LV only on SSD-tier PVs
lvcreate -L 50G -n lv_fast vg_mixed @tier_ssd
 
# Create LV only on HDD-tier PVs
lvcreate -L 500G -n lv_archive vg_mixed @tier_hdd
 
# Extend LV using only same-tier storage (cling_by_tags)
lvextend --alloc cling_by_tags -L +25G /dev/vg_mixed/lv_fast
 
# Create mirrored LV with legs on different racks (fault isolation)
lvcreate --type raid1 -L 100G -n lv_critical --alloc anywhere \
    vg_mixed @rack_a @rack_b
 
# Remove a tag
pvchange --deltag @tier_ssd /dev/nvme0n1p2
 
# List all available tags in system
vgs -o tags
pvs -o tags

Strategic Tagging

Volume Group Operations

Beyond creation and extension, volume groups support various administrative operations for maintenance, migration, and troubleshooting.

Activation and Deactivation:

Volume groups and their logical volumes must be activated before use. Activation loads the device-mapper mappings into the kernel, making the LVs accessible as block devices.

VG Administrative Operations
Bash
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
#!/bin/bash
# Volume Group Administrative Operations
 
# === ACTIVATION CONTROL ===
 
# Activate all LVs in a VG
vgchange -ay vg_data
 
# Deactivate all LVs in a VG (must unmount first!)
vgchange -an vg_data
 
# Activate specific LV only
lvchange -ay /dev/vg_data/lv_database
 
# Check activation status
lvs -o lv_name,lv_attr vg_data
# Attribute field 5: 'a' = active, '-' = inactive
 
# Set auto-activation on boot
vgchange --setautoactivation y vg_data
 
# Disable auto-activation (for manual control)
vgchange --setautoactivation n vg_data
 
# === VG RENAME ===
 
# Rename volume group (all LVs automatically get new paths)
vgrename vg_data vg_storage
 
# Verify new name
vgs vg_storage
 
# Update fstab entries for new device paths!
 
# === VG EXPORT/IMPORT (for moving between systems) ===
 
# Step 1: Deactivate VG
vgchange -an vg_portable
 
# Step 2: Export VG (marks it as foreign)
vgexport vg_portable
 
# === On new system ===
 
# Step 3: Scan for all PVs
pvscan
 
# Step 4: Import the VG
vgimport vg_portable
 
# Step 5: Activate
vgchange -ay vg_portable
 
# === VG SPLIT (separate PVs into new VG) ===
 
# Move LV and its PVs to new VG (must specify all necessary PVs)
vgsplit vg_data vg_archive /dev/sdd /dev/sde -n lv_old_data
 
# Note: Source VG must have no LVs with extents on moved PVs
# except for the ones being moved
 
# === VG MERGE (combine VGs) ===
 
# Merge vg_archive into vg_data
# Deactivate source VG first
vgchange -an vg_archive
vgmerge vg_data vg_archive
 
# === VG ATTRIBUTE CHANGES ===
 
# Make VG read-only (all LVs become read-only)
vgchange -p r vg_archive
 
# Restore read-write
vgchange -p rw vg_archive
 
# Set VG maximum LV limit
vgchange -l 50 vg_data
 
# Remove the limit (set to 0 = unlimited)
vgchange -l 0 vg_data
 
# === RECOVERY OPERATIONS ===
 
# Repair VG with missing PVs
vgreduce --removemissing --force vg_data
 
# Restore VG from metadata backup
vgcfgrestore -f /backup/vg_data.txt vg_data
 
# Rebuild metadata from disk scan
vgscan --mknodes
 
# Force reread of all LVM metadata
pvscan --cache

VG Export/Import Workflow:

When moving storage between systems (e.g., moving a disk array to a new server), the export/import workflow ensures clean handoff:

Deactivate: Ensures no I/O is in progress
Export: Marks metadata indicating VG is in transit
Physical move: Relocate the disks/LUNs
Scan: New system discovers the PVs
Import: Clears the export flag, registers VG on new system
Activate: Makes LVs accessible

This workflow prevents accidentally accessing the same storage from multiple systems, which could cause data corruption.

VG Merge Prerequisites

Production Best Practices

Design Recommendations

•Separate VGs by failure domain: Don't mix local disks and SAN storage in one VG; failures affect differently
•Separate VGs by performance tier: Keep SSDs and HDDs in separate VGs for predictable performance
•Use descriptive VG names: Names like 'vg_prod_db_ssd' are self-documenting
•Reserve 15-20% free space: Enables snapshots, emergencies, and pvmove operations
•Prefer fewer, larger VGs: Reduces administrative overhead; use tags for allocation control
•Document extent size rationale: Future administrators need to understand why 8MB vs 4MB was chosen

Common Mistakes to Avoid

•Allocating 100% of VG space: Leaves no room for snapshots or growth
•Mixing boot and data VGs: Boot issues shouldn't affect data availability
•Using random extent sizes: Stick to defaults unless specific requirements exist
•Ignoring metadata redundancy: At least 2 PVs should have metadata copies
•No external metadata backup: Losing all PVs means losing everything
•Complex VG topologies for small deployments: Adds complexity without benefit

Monitoring and Capacity Planning:

Proactive monitoring prevents emergency situations:

VG Monitoring Script
Bash
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
#!/bin/bash
# LVM Monitoring Script - Add to cron for regular checks
 
LOG="/var/log/lvm_monitor.log"
THRESHOLD=85  # Alert when VG is 85% full
 
echo "=== LVM Health Check: $(date) ===" >> $LOG
 
# Check VG capacity
for vg in $(vgs --noheadings -o vg_name); do
    total=$(vgs --noheadings --units g -o vg_size $vg | tr -d ' g')
    free=$(vgs --noheadings --units g -o vg_free $vg | tr -d ' g')
    used_pct=$(echo "scale=0; (($total - $free) / $total) * 100" | bc)
    
    if [ $used_pct -ge $THRESHOLD ]; then
        echo "WARNING: VG $vg is ${used_pct}% full" >> $LOG
        # Send alert (email, Slack, PagerDuty, etc.)
        # mail -s "LVM Alert: $vg at ${used_pct}%" admin@example.com
    fi
done
 
# Check for missing PVs
pvscan 2>&1 | grep -i "missing" >> $LOG
 
# Check metadata sequence consistency
for vg in $(vgs --noheadings -o vg_name); do
    vgck $vg 2>> $LOG
done
 
# List PVs with low free space (individual PV capacity)
pvs --noheadings -o pv_name,pv_size,pv_free | while read pv size free; do
    # Alert if individual PV is nearly full (affects allocation locality)
    # This is informational for cling policy effectiveness
    echo "PV: $pv  Size: $size  Free: $free" >> $LOG
done

Documentation Habit

Summary: Volume Groups

Volume Groups form the aggregation layer of LVM, pooling physical storage into flexible capacity pools. Let's consolidate the essential concepts covered:

Key Takeaways

•Extent size is permanent: Choose carefully at VG creation; 4MB default suits most workloads
•Metadata redundancy is automatic: Every PV stores VG metadata by default, providing resilience
•VGs can grow dynamically: Adding PVs with vgextend immediately makes capacity available
•PV removal requires data migration: Use pvmove before vgreduce to safely remove hardware
•Allocation policies control placement: From contiguous (performance) to anywhere (flexibility)
•Tags enable sophisticated control: Label PVs by tier, location, or purpose for targeted allocation
•Export/Import enables portability: Move VGs between systems cleanly and safely
•Reserve free space: 15-20% unallocated enables snapshots and emergency capacity

What's Next:

Page Complete

2 / 5