Before proposing another way to back up, we needed to see where the night of a cPanel server with hundreds of accounts really goes. What does JetBackup 5 do, step by step, during a cPanel server’s nightly backup?
371accounts in 2h10, up to 5 in parallel (4 or 5 for 88% of the time)
87%of the time is transfer; 13% is preparation, almost all of it database dumps
Why it matters to whoever runs the server
- A 2h10 night for 371 accounts is the window in which the sites share the server with the backup and the storage takes everything. The longer the window, the likelier it runs into business hours.
- 87% of the time is transfer: that’s where a new backup has to win, sending less data and giving the storage less work. The 13% of preparation is almost all database dumps.
- Knowing what failed, and where, is what prevents the next-day ticket: that’s why Salvo’s status says which mailbox, which database, which table.
Median
47 s
90% of accounts within
190 s
Slowest account
19.5 min
Source: passive observation, JetBackup 5.4.1.4 · Oct 13, 2026
See the data as a table
| What we saw | Value | Detail |
|---|---|---|
| Accounts that night | 371 | 2h10, up to 5 in parallel |
| CPU limit | CPUQuota=200% | via systemd; load around 11 |
| Final status | partial | 17 accounts: 16 with mailboxes without a password (143 mailboxes, normal), 1 with a real database error |
| Database dump | --skip-lock-tables | without --single-transaction |
| Dump before upload | .sql.gz | written to local disk |
| 10 slowest accounts | 22% | of the night’s time |
Method
Passive observation of a production cPanel server with about 370 accounts, JetBackup 5.4.1.4.
For 2h10, every 30 seconds, we recorded processes, open files, disk and network usage: 266 samples.
Nothing was written to the server and no process was traced with strace.
What we decided
Salvo’s status says what failed (which mailbox, database or table) and tells a normal state apart from a real error. And MySQL dumps use --single-transaction: a dump that doesn’t finish is treated as a failure, not as a backup.
What we don’t know yetThis is one night on one server. Other JetBackup setups (destinations, schedules, limits) may behave differently. Samples every 30 s miss short events; per-account times carry that margin.
Want to see Salvo on your server?Join the beta waitlist. We’ll let you know when there’s a spot.
How it connects to the other labs
Less work (IOPS) for the storage
1One new hardlink per file, every night: why backup storage saturatesLab 04 · Oct 14, 2026~90Writing to the storage disk was as fast as throwing the data awayLab 05 · Oct 15, 2026166×The first backup side by side: 45% less time and 166× fewer files on the storageLab 14 · Oct 16, 2026152×The incremental right after: 152× fewer reads on the storage diskLab 15 · Oct 16, 2026