Snapshot vs File-Based Backups for VPS Servers

Backups for a VPS come in two families, and choosing between them is a speed-versus-granularity tradeoff you should make deliberately rather than accidentally. Snapshot-based backups act at the block level and capture whole disks in seconds. File-based backups walk the filesystem and produce dumps or archives you can browse, grep, and restore one piece at a time. Both are legitimate; each has blind spots.

How snapshot backups behave

A host-level snapshot copies the virtual disk at a point in time using instant copy mechanics rather than a byte-by-byte pass. The snapshot appears in seconds and mounts in about the time it takes to attach a disk. Databases are the catch: a crash-consistent snapshot of a database caught mid-write can require recovery work at boot. The fix is to quiesce first, snapshot, then release.

# quiesce MySQL/MariaDB around a host-level snapshot
mysql -e 'FLUSH TABLES WITH READ LOCK;'
# trigger the snapshot at your provider layer here
mysql -e 'UNLOCK TABLES;'

Snapshots win on frequency and speed but are coarse on granularity: restoring one deleted file means mounting a whole disk image.

How file-based backups behave

A logical dump or archive produced by pg_dump, mysqldump, or tar is slower but vastly more inspectable. You can restore one table, one directory, or one row. File-based backups also ride along with the consistency handling your database tool provides, and they are the natural input for an off-site object store copy.

# file-based: portable and granular
pg_dump --format=custom --compress=9 appdb   | rclone rcat OBJECTSTORE:backups/appdb.dump
tar -czf /var/backups/www.20260801.tar.gz -C /srv www

rclone rcat reads the archive from stdin, so a dump never touches a temporary file.

Decide with a short matrix

• Frequent small writes, tight RPO: snapshots every few hours plus a nightly logical dump for portability.
• App and static files, 24-hour RPO: one tar per night is enough, with a snapshot before OS updates.
• Compliance or migration needs: logical dumps, because they cross versions and vendors.
• Ransomware or accounting needs: snapshots only help if off-site copies ride along.

The winning pattern is usually both: a cheap frequent layer plus a portable infrequent layer, instead of an argument about which single tool is superior.

Cost and retention of each

A snapshot's storage footprint grows with the blocks that changed, not the whole disk, so dozens of cheap snapshots stay small when churn is low. File archives grow by whole-file size per run unless the tool deduplicates. Over 30 days both settle under a few hundred GB for a typical 10 GB database server; snapshots stay a bit cheaper, files stay a lot easier to restore from.

The hybrid that actually works

Nightly logical dumps go off-site to object storage; hourly host-level snapshots protect against configuration drift and give a fast rollback for bad deploys. Within a year this is the pattern most production VPS installations settle into.

Takeaway

Snapshot backups buy speed, file-based backups buy granularity and portability, and a serious setup uses both. Quiesce databases before a snapshot, keep logical dumps as the off-site layer, and schedule each on its own cadence. A Netbay VPS in Lucknow DC01 gives you the flexibility to run either pattern — start one at netbayhosts.in.