<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:content="http://purl.org/rss/1.0/modules/content/"><channel><title>Btrfs on rHomelab</title><link>https://rhomelab.com/tags/btrfs/</link><description>Recent content in Btrfs on rHomelab</description><generator>Hugo</generator><language>en-US</language><lastBuildDate>Fri, 20 Nov 2026 00:00:00 +0000</lastBuildDate><atom:link href="https://rhomelab.com/tags/btrfs/index.xml" rel="self" type="application/rss+xml"/><item><title>Btrfs for Data Hoarders: Subvolumes, Snapshots, and When It Actually Beats ZFS</title><link>https://rhomelab.com/data-hoarding/btrfs-for-data-hoarders/</link><pubDate>Fri, 20 Nov 2026 00:00:00 +0000</pubDate><guid>https://rhomelab.com/data-hoarding/btrfs-for-data-hoarders/</guid><description>ZFS gets most of the attention on this site, but Btrfs is a real, mature alternative with its own strengths. Here&amp;#39;s what it actually offers a data hoarder, and the cases where it&amp;#39;s the better pick.</description><content:encoded><![CDATA[<p>Every ZFS article on this site mentions Btrfs in passing as the &ldquo;other&rdquo; copy-on-write filesystem, usually in a single sentence before moving on. That&rsquo;s fair when the topic is ZFS specifically, but it undersells Btrfs on its own terms. It&rsquo;s a mature, in-kernel filesystem with checksummed data, instant snapshots, transparent compression, and its own answer to RAID, and for a meaningful slice of data hoarders it&rsquo;s the better default, not just the fallback option you settle for when ZFS isn&rsquo;t available.</p>
<p>This isn&rsquo;t a &ldquo;Btrfs beats ZFS&rdquo; argument. It&rsquo;s a look at what Btrfs actually gives you, where it genuinely falls short, and the specific situations where picking it over ZFS is the right call rather than a compromise.</p>
<h2 id="the-one-structural-difference-that-drives-everything-else">The one structural difference that drives everything else</h2>
<p>ZFS (specifically OpenZFS, the version anyone actually runs) is not part of the Linux kernel. It ships as an out-of-tree module, built via DKMS against whatever kernel you&rsquo;re running, because of a long-standing license incompatibility between ZFS&rsquo;s CDDL license and the kernel&rsquo;s GPL. In practice this means a kernel update can require a module rebuild before ZFS comes back online, and some distributions (Fedora being the clearest example) don&rsquo;t ship it at all, leaving you to add a third-party repo and accept the maintenance burden that comes with it.</p>
<p>Btrfs has none of that friction, because it&rsquo;s mainlined into the kernel itself. Install any mainstream Linux distribution with a reasonably current kernel and Btrfs is already there, no module, no DKMS, no license dance. openSUSE and Fedora both ship it as an installer option, and it&rsquo;s the filesystem underneath Synology&rsquo;s Hybrid RAID on newer DSM releases. If you&rsquo;ve ever had a Proxmox host or a homelab box sit unbootable after a kernel update because the ZFS module didn&rsquo;t rebuild cleanly, that&rsquo;s a failure mode Btrfs structurally can&rsquo;t have.</p>
<h2 id="subvolumes-lighter-than-zfs-datasets-less-isolated-too">Subvolumes: lighter than ZFS datasets, less isolated too</h2>
<p>A Btrfs subvolume is roughly ZFS&rsquo;s dataset concept, a separately snapshottable, separately mountable namespace living inside one filesystem, but it&rsquo;s a lighter-weight abstraction. Creating one is instant and cheap:</p>
<div class="highlight"><pre tabindex="0" class="chroma"><code class="language-bash" data-lang="bash"><span class="line"><span class="cl">btrfs subvolume create /mnt/tank/media
</span></span><span class="line"><span class="cl">btrfs subvolume create /mnt/tank/documents
</span></span><span class="line"><span class="cl">btrfs subvolume list /mnt/tank
</span></span></code></pre></div><p>The tradeoff is that subvolumes share the parent filesystem&rsquo;s free space pool entirely, with no built-in per-dataset space accounting unless you layer on quota groups (<code>qgroup</code>), which carry a real performance cost and a reputation for being finicky at scale. ZFS datasets get per-dataset quotas and reservations essentially for free. If you need hard space guarantees between different data sets on the same pool, ZFS&rsquo;s model is the more battle-tested one.</p>
<h2 id="snapshots-and-replication">Snapshots and replication</h2>
<p>Btrfs snapshots work on the same copy-on-write principle as ZFS: a snapshot is a reference to the current state of a subvolume&rsquo;s blocks, instant to create regardless of size, and it costs nothing in space until the live data diverges from it.</p>
<div class="highlight"><pre tabindex="0" class="chroma"><code class="language-bash" data-lang="bash"><span class="line"><span class="cl"><span class="c1"># snapshot a subvolume</span>
</span></span><span class="line"><span class="cl">btrfs subvolume snapshot /mnt/tank/media /mnt/tank/.snapshots/media-2026-11-20
</span></span><span class="line"><span class="cl">
</span></span><span class="line"><span class="cl"><span class="c1"># browse it like any other directory, no special mount needed</span>
</span></span><span class="line"><span class="cl">ls /mnt/tank/.snapshots/media-2026-11-20
</span></span><span class="line"><span class="cl">
</span></span><span class="line"><span class="cl"><span class="c1"># send a snapshot to another filesystem or host</span>
</span></span><span class="line"><span class="cl">btrfs send /mnt/tank/.snapshots/media-2026-11-20 <span class="p">|</span> btrfs receive /mnt/backup/
</span></span></code></pre></div><p>The practical difference from ZFS&rsquo;s <code>.zfs/snapshot/</code> directory is cosmetic. Btrfs snapshots are just subvolumes themselves, sitting wherever you created them, no hidden path to remember. For scheduling and retention, <strong>snapper</strong> is the Btrfs-world equivalent of ZFS&rsquo;s <code>sanoid</code>, handling automatic snapshot creation and pruning against a policy you define per subvolume, rather than hand-writing cron jobs:</p>
<pre tabindex="0"><code># /etc/snapper/configs/media
TIMELINE_CREATE=&#34;yes&#34;
TIMELINE_LIMIT_HOURLY=&#34;24&#34;
TIMELINE_LIMIT_DAILY=&#34;14&#34;
TIMELINE_LIMIT_MONTHLY=&#34;3&#34;
</code></pre><p>Everything written about <a href="/data-hoarding/zfs-snapshots-for-data-hoarders/">ZFS snapshots as ransomware and mistake protection</a> applies here unchanged: snapshots aren&rsquo;t a backup, they live on the same disks as the data they protect, and the account that writes to your share should never have permission to delete them. <code>btrfs send</code>/<code>receive</code> for off-box replication is the direct equivalent of <code>zfs send</code>/<code>receive</code>, and the same &ldquo;send snapshots somewhere the primary host can&rsquo;t delete them&rdquo; logic holds for real ransomware protection.</p>
<h2 id="checksums-scrubbing-and-the-dedup-gap">Checksums, scrubbing, and the dedup gap</h2>
<p>Btrfs checksums both data and metadata by default, and <code>btrfs scrub</code> walks the filesystem verifying every checksum against its block, repairing anything that doesn&rsquo;t match from a surviving redundant copy, exactly like a ZFS scrub. The <a href="/data-hoarding/bit-rot-checksums-scrubbing/">bit rot and checksumming article</a> already covers why this matters more than RAID redundancy alone, and that logic carries over directly, a Btrfs volume with corruption and no redundant copy to heal from will tell you about the problem, which plain ext4 or XFS never will.</p>
<p>Where Btrfs genuinely falls short of ZFS is block-level deduplication. ZFS has it built in (expensive in RAM, rarely worth turning on, but it exists). Btrfs has no online dedup at all, only offline tools like <code>duperemove</code> that scan after the fact and use reflinks to collapse duplicate blocks, covered in more depth in <a href="/data-hoarding/finding-duplicate-files-dedup-tools/">finding and removing duplicate files</a>. For a data hoarder whose dedup needs are mostly &ldquo;stop me from keeping three copies of the same file,&rdquo; offline dedup is plenty. For a workload that actually benefits from continuous block-level dedup, ZFS is the only one of the two that offers it.</p>
<h2 id="the-raid-question-where-btrfs-actually-loses">The RAID question: where Btrfs actually loses</h2>
<p>This is the one area where the usual one-sentence caveat in other articles on this site deserves real elaboration. Btrfs&rsquo;s RAID0, RAID1, and RAID10 profiles are solid, production-safe, and have been for years, mirroring or striping across devices with the same checksummed self-healing as everything else in the filesystem.</p>
<p>Btrfs&rsquo;s RAID5 and RAID6 profiles are a different story. They have a long-documented &ldquo;write hole&rdquo; problem, a window where a crash or power loss during a partial stripe write can corrupt both the data and its parity simultaneously, and the Btrfs wiki itself has carried warnings against using them in production for years. As of 2026 this still hasn&rsquo;t been fully resolved upstream. If you want parity RAID with Btrfs&rsquo;s checksumming and snapshot model, don&rsquo;t use Btrfs&rsquo;s own RAID5/6, mirror-only is the safe profile. ZFS&rsquo;s RAIDZ has no equivalent write-hole issue, and that gap alone is reason enough for anyone building a large parity array to default to ZFS over Btrfs for that specific job.</p>
<h2 id="where-btrfs-is-genuinely-the-better-pick">Where Btrfs is genuinely the better pick</h2>
<ul>
<li><strong>You&rsquo;re already on a Btrfs-native distro or NAS OS.</strong> Synology&rsquo;s Btrfs-based Hybrid RAID, openSUSE, or any setup where Btrfs is already the root filesystem. Adding ZFS on top means a second filesystem&rsquo;s worth of tooling, a kernel module dependency, and a separate mental model, for a benefit (mostly RAIDZ and block dedup) you may not need.</li>
<li><strong>You want to avoid the DKMS/kernel-update dependency entirely.</strong> Especially relevant on a distro with a fast-moving kernel (Fedora, Arch-based setups) where &ldquo;will ZFS still boot after I update&rdquo; is a recurring, avoidable question.</li>
<li><strong>You&rsquo;re growing a pool by adding one mismatched drive at a time, not replacing a full vdev set.</strong> Btrfs&rsquo;s <code>device add</code> plus <code>btrfs balance</code> lets you fold in drives of any size one at a time without planning around matched vdev members. ZFS&rsquo;s raidz expansion has improved in recent OpenZFS releases, but Btrfs has supported this incremental, mismatched-size growth pattern natively for far longer.</li>
<li><strong>Your redundancy need is mirroring, not parity.</strong> RAID1/RAID10 on Btrfs is as safe as ZFS&rsquo;s mirror vdevs, with the licensing and kernel-integration advantages on top.</li>
</ul>
<h2 id="where-zfs-is-still-the-right-call">Where ZFS is still the right call</h2>
<p>If you&rsquo;re building a large parity array from scratch, want per-dataset quotas and reservations, need real block-level dedup, or just want the single most battle-tested, widely-documented option for a serious data hoard, ZFS remains the safer default, and it&rsquo;s why it gets the lion&rsquo;s share of coverage on this site. Btrfs isn&rsquo;t a downgrade from ZFS. It&rsquo;s a different filesystem with a different set of tradeoffs, and for the specific cases above, it&rsquo;s the one that actually fits better.</p>
]]></content:encoded></item></channel></rss>