- SnapRAID CLI v14.10 (Released July 2026). + SnapRAID CLI v15.0 (Released October 2026).
- SnapRAID Daemon v1.15 (Released July 2026). + SnapRAID Daemon v2.0 (Released October 2026).
@@ -112,39 +112,39 @@
-SnapRAID creates a file-system snapshot conceptually similar to other snapshot solutions. -However, its primary purpose is to recover a disk to its saved state after a complete disk failure. +SnapRAID provides array-level parity protection across multiple independent disks, whereas filesystem snapshots only protect against accidental modifications or deletions within a single filesystem, without protecting against physical disk failure.
-SnapRAID is more akin to the RAID-Z/RAID functionality of ZFS/Btrfs. -Compared to them, SnapRAID offers the following advantages: +Rather than being mutually exclusive, SnapRAID can directly integrate with them: with the snapshot option enabled, SnapRAID automatically creates and manages snapshots on Btrfs, ZFS, Bcachefs, and NTFS. This combines the point-in-time consistency and modification protection of local filesystem snapshots with SnapRAID's multi-disk parity recovery. +
++Compared to the RAID capabilities of ZFS (RAID-Z) and Btrfs (RAID-5/6), SnapRAID offers several distinct advantages:
-ZFS and Btrfs have advantages, such as being real-time solutions. -The best choice depends on your specific use case. +ZFS and Btrfs have advantages as well, such as real-time parity updates. The best choice depends on your specific use case.
Yes, SnapRAID Daemon includes a web GUI. -There is also an external GUI called - -Elucidate - -, -and a plugin for +There are also plugins for OpenMediaVault +and + +Cockpit + .
-For a setup example in Ubuntu 14.04, see +For a setup example in Ubuntu, see Zack's SnapRAID tutorial @@ -565,20 +564,26 @@ Result
+To enable filesystem snapshots, add the + +snapshot + +option to your + +snapraid.conf + +configuration file: +
++snapshot ++
+This option applies to data disks formatted with Btrfs, Bcachefs, or ZFS in Linux, and NTFS in Windows. +
++The use of snapshots is completely automatic and transparent to both the user and the SnapRAID daemon. +No changes are required to your workflow, commands, or file management. SnapRAID automatically creates, uses, and cleans up snapshots in the background during operations (such as sync, scrub, check, and fix). +
++The only visible effect is informational progress messages displayed during operations, such as: +
++Creating disk d1 scan snapshot... +Committing disk d1 stable snapshot... +Using disk d1 stable snapshot... ++
+Note that snapshot creation and deletion require administrative privileges. Ensure SnapRAID is run with the necessary permissions (e.g., sudo on Linux or as Administrator on Windows). +
+-For optimal performance, connect all disks via SATA, not USB. +For optimal performance, connect all disks via SATA, SAS, or NVMe, and avoid USB disks when possible. +
++Both traditional mechanical hard drives (HDDs) and solid-state drives (SATA or NVMe SSDs) are supported. +SnapRAID automatically detects non-rotational devices on both Linux and Windows, skipping unnecessary spin-up and spin-down operations.
See ServerBuilds NAS Killer 4.0 Build Guide, Reddit DataHoarder and -Reddit HomeServer. +Reddit HomeServer for examples of hardware setups.
-No. SnapRAID is typically not CPU-bound; the limiting factor is disk read performance. -A single-core CPU is usually sufficient. -
--For example, in my system, I achieve a combined read speed of 600 MiB/s, -but SnapRAID can hash at 12,000 MiB/s on a single core. +SnapRAID is typically not CPU-bound; the limiting factor is disk read and write performance. +Even an entry-level modern CPU is usually sufficient, as SnapRAID's hashing and parity computation algorithms are highly optimized and can process data much faster than mechanical disks can deliver it.
Use - + Volume Mount Points to mount a disk in a directory without @@ -1089,14 +1125,14 @@ Faulty RAM is the most common cause of data loss with SnapRAID. Run the "sync" command frequently, from once a day to once a week.
-To safely move files between data disks in the array, follow these steps: +If the snapshot option is enabled in snapraid.conf, you can move files directly from one disk to another and run sync whenever convenient. +Because the original files remain preserved in the filesystem snapshot, parity protection is not compromised during the process. +
++If filesystem snapshots are not enabled, follow these steps to move files safely between data disks:
-In Linux, use the +You can limit SnapRAID's disk I/O throughput on any platform using the built-in -w, --bw-limit RATE option. +This sets a global bandwidth limit across all disks (e.g. 100M for 100 MB/s): +
++snapraid -w 100M sync ++
+Additionally, in Linux, you can use the ionice -command with idle priority, like: +command with idle priority:
ionice -c 3 snapraid sync @@ -1426,7 +1473,11 @@ What happens if I delete files and a disk fails before updating parity with "syn-In the worst case, files deleted or modified on non-failed disks may prevent recovery of an equivalent amount of data on the failed disk. +If the snapshot option is enabled in snapraid.conf, this risk is completely eliminated. +Because deleted or modified files are preserved in filesystem snapshots, SnapRAID can access the original data blocks and recover the failed disk without any data loss. +
++If filesystem snapshots are not enabled, in the worst case, files deleted or modified on non-failed disks may prevent recovery of an equivalent amount of data on the failed disk. For example, deleting 10 GB of data may prevent recovery of 10 GB from the failed disk. The exact data loss depends on how much the deleted and failed data overlap in the parity.
@@ -1434,7 +1485,7 @@ The exact data loss depends on how much the deleted and failed data overlap in t Using two parity disks significantly improves recovery chances.-To ensure safety, avoid deleting files directly. Instead, move them to another directory on the same disk, +Without snapshots, to ensure safety, avoid deleting files directly. Instead, move them to another directory on the same disk, outside the SnapRAID-checked directory tree. Run the "sync" command, then delete the files. If a disk fails during the "sync", run the "fix" command, specifying the location of the deleted files:
@@ -1451,7 +1502,13 @@ What happens if a disk fails during a "sync"?You can still recover data. In the worst case, you can recover as much data as if the disk failed before the "sync". If the "sync" process has progressed, SnapRAID uses the partially synced data to recover more. -Using the "autosave" configuration option saves intermediate content files during the sync process, improving recovery. +
++SnapRAID preserves hash data in the content file across interruptions to ensure reliable recovery, and the --import option remains functional even after an incomplete sync. +Enabling the snapshot option further improves recovery by maintaining access to both the previous stable snapshot and the current progress. +
++Using the "autosave" configuration option also saves intermediate content files during long syncs, further improving recovery.
-Using double parity improves recovery when there are overlapping changes in the array. +This is possible because SnapRAID stores the expected hash of each data block: after reconstructing a block from parity, it verifies it against the stored hash to ensure that the recovered data is correct. +
++When the snapshot option is enabled, SnapRAID can read automatically from both the previous stable snapshot and the pending snapshot. This maximizes the probability of a full recovery by providing access to the exact data blocks required to solve the parity equations, even if those blocks were modified or deleted between syncs. +
++Using double parity also improves recovery when there are overlapping changes in the array.
-You can also write, and if this interferes with the 'sync', -the process continues, skipping the written part. +If the snapshot option is enabled, you can also write to and modify files without any interference: SnapRAID reads from a frozen, read-only snapshot created at the start of the operation, ensuring that live filesystem modifications do not cause parity mismatches or aborted operations.
-This may affect potential recovery, but it is equivalent to writing just after the 'sync' completes. +Without snapshots, you can still write, but if a file being read is concurrently modified, the process skips the written part. This may affect potential recovery until the next sync, which is equivalent to modifying the file immediately after the sync completes.
@@ -1723,7 +1785,7 @@ What is the optimal hardware configuration for the best performance?-The most critical factor is using (E)SATA connections for all disks to avoid controller slowdowns. +The most critical factor is using internal connections (SATA, SAS, or NVMe) for all disks to avoid controller slowdowns. Avoid USB disks.
@@ -1809,8 +1871,8 @@ How fast are the hash and parity computations?
Murmur3 hash is very fast on 32-bit systems, and SpookyHash is even faster on 64-bit systems. -Parity computation is efficient up to double parity, even without SSE support. -Beyond double parity, a CPU with SSSE3 support is recommended. +Parity computation is heavily optimized using hardware SIMD acceleration across all major architectures: SSE2, SSSE3, AVX2, and AVX-512 on x86/x64, as well as ARM NEON on 64-bit (AArch64) and 32-bit (AArch32) ARM platforms. +Parity computation is efficient up to double parity even on basic CPUs. Beyond double parity, a CPU with SSSE3, AVX2, AVX-512, or NEON support is recommended for optimal performance. For low-end CPUs without SSE, a special, incompatible triple-parity option is available. See 'z-parity' in the manual for details.
@@ -1820,11 +1882,19 @@ Check your values with the -T option:
root@redstar:/root# snapraid -T
-snapraid v14.0 by Andrea Mazzoleni, https://www.snapraid.it
+snapraid v15.0 by Andrea Mazzoleni, https://www.snapraid.it
Compiler gcc 11.2.0
-CPU GenuineIntel, family 6, model 165, flags sse2 ssse3 crc32 avx2
-Memory is little-endian 64-bit
+CPU GenuineIntel, family 6 (0x6), model 165 (0xa5), stepping 5, flags sse2 ssse3 crc32 avx2
+Model Intel(R) Core(TM) i7-10700 CPU @ 2.90GHz
+Clock 2.90 GHz (max 4.80 GHz)
+Cache L1 Data 32 KB, L2 256 KB, L3 16384 KB
+Machine has 16 cores
+Running on cpu 1
+Memory 31806 MB, little-endian 64-bit, page 4 KB
+128-bit integers are supported
Support nanosecond timestamps with futimens()
+Host Slackware 15.0 x86_64 (post 15.0 -current) (Linux 6.18.38)
+System Dell Inc. OptiPlex 5080
Speed test using 8 data buffers of 262144 bytes, for a total of 2048 KiB.
Memory blocks have a displacement of 1792 bytes to improve cache performance.
@@ -1832,43 +1902,88 @@ The reported values are the aggregate bandwidth of all data blocks in MB/s,
not counting parity blocks.
Memory write speed using the C memset() function:
- memset 61338
+ memset 58052
CRC used to check the content file integrity:
- table 1665
- intel 12488
+ table 1669
+ intel 36099
Hash used to check the data blocks integrity:
- best murmur3 spooky2 metro museair
- hash spooky2 6172 18024 20752 29206
+ best murmur3 spooky2 museair
+ hash spooky2 6172 17914 30727
-RAID functions used for computing the parity with 'sync':
+RAID polynomial functions used for computing parity with 'sync':
best int8 int32 int64 sse2 sse2e ssse3 ssse3e avx2 avx2e
- gen1 avx2 19147 34378 62743 78638
- gen2 avx2 5303 9850 26288 26926 45010
- genz avx2e 3086 4950 14319 13838 25299
- gen3 avx2e 1278 13042 14120 27155
- gen4 avx2e 960 9895 11008 21764
- gen5 avx2e 784 7960 8713 17174
- gen6 avx2e 652 6502 7502 14638
+ gen1 avx2 20738 37638 60622 70014
+ gen2 avx2e 2465 5220 9863 27038 30954 45964 51622
+ gen3 avx2e 1659 12945 14665 28876
+ gen4 avx2e 1142 9284 10311 22265
+ gen5 avx2e 862 7206 7796 14847
+ gen6 avx2e 673 5829 6487 12501
-RAID functions used for recovering with 'fix':
- best int8 ssse3 avx2
- rec1 avx2 1602 3942 4301
- rec2 avx2 698 1731 2224
- rec3 avx2 166 908 1391
- rec4 avx2 108 607 973
- rec5 avx2 76 421 694
- rec6 avx2 57 313 532
+AES polynomial functions used for computing parity with 'sync':
+ best int8 int32 int64 sse2 sse2e ssse3 ssse3e avx2 avx2e
+ gen1 avx2 20712 37765 60677 70051
+ gen2 avx2e 2465 4763 8456 23497 28220 41045 49308
+ gen3 avx2e 1659 12043 13606 27257
+ gen4 avx2e 1137 8825 9945 21277
+ gen5 avx2e 857 7543 14513
+ gen6 avx2e 663 6335 12264
+
+Vandermonde functions used for computing parity with 'sync':
+ best int8 int32 int64 sse2 sse2e ssse3 ssse3e avx2 avx2e
+ genz avx2e 2918 5332 14488 14450 25930
+
+RAID polynomial functions used for recovering with 'fix':
+ best int ssse3 ssse3e avx2 avx2e
+ rec1of1 avx2e 37680 58508 58457 66444 66444
+ rec1of2 avx2e 2896 26525 28901 44590 46896
+ rec1of3 avx2e 2898 18517 20460 37638 37638
+ rec2of2 avx2e 2585 19611 21592 27346 31388
+ rec2of3 avx2e 1749 12875 13389 19440 23984
+ rec2of4 avx2e 1744 12865 13336 19228 23933
+ rec3of3 avx2e 1667 9436 10658 18325 20636
+ rec4of4 avx2e 1154 6584 8422 12700 15033
+ rec5of5 avx2e 853 4355 7025 8648 12372
+ rec6of6 avx2e 692 3682 5370 7313 10466
+
+(recNofM: N data disks recovered using parities up to the M-th parity,
+ e.g. rec1of1 uses P, rec1of2 uses Q, rec2of2 uses P+Q, rec2of3 uses Q+R)
+SnapRAID Daemon monitors disk health and warns whenever a SMART attribute marked as PRE-FAIL degrades its normalized value, allowing you to anticipate potential drive issues early. +
++However, certain hard drive models exhibit natural fluctuations in non-critical attributes during normal operation, most commonly Spin_Up_Time (attribute 3) or Raw_Read_Error_Rate (attribute 1). +These normal variations can cause harmless warnings on the WebUI dashboard and trigger notification emails. +
++You can silence these noisy attributes in the daemon configuration file (snapraidd.conf) using the sys_smartignore option. +Specify the disk name (or * for all disks) followed by attribute names or numbers: +
++sys_smartignore = * Spin_Up_Time +sys_smartignore = * Spin_Up_Time Raw_Read_Error_Rate +sys_smartignore = parity 188 Command_Timeout ++
+Even when an attribute is suppressed with sys_smartignore, your disks remain fully protected: if an attribute exceeds the manufacturer's critical failure threshold, SnapRAID Daemon will always alert you. +
+The '-T largefile4' and '-m 0' options provide more space for the parity file.
++You can also split the parity across multiple disks or partitions using split parity. +In the configuration file, you can specify multiple comma-separated files for the same parity level (up to 8 files): +
++parity /mnt/diskp1/snapraid.parity,/mnt/diskp2/snapraid.parity ++
+As soon as the first file cannot grow anymore due to lack of space, SnapRAID automatically starts growing the next file in the list. +