All work

Project 02VPS · HestiaCP · Web & mail · 2026

Servers and mail that just keep working

Websites, apps, databases and company email for three businesses, including Leqta and Bach Saiss, all on one HestiaCP server. I moved everything to a new cloud VPS without losing a single email, then kept it healthy, secure and easy for each business to manage.

Clients
Leqta, Bach Saiss and one more
What I did
Audit, migration, mail, security, upkeep
Platform
HestiaCP on Ubuntu 24.04
Cutover
23 October 2026, one afternoon
12websites
7Node.js apps under PM2
3databases, checked table by table
29mailboxes on 3 domains
3,683emails moved, 0 lost
[01] // THE_MIGRATION

Six steps, and nothing switched until everything matched

The old server was ageing and nobody had a full picture of what lived on it. The plan was simple to say and careful to run: copy everything, prove the copy is identical, then switch.

STEP_01 · Audit

Find what the backups miss

The panel's own backup left out the app code, the PostgreSQL database, uploads, settings files and custom web templates. For two sites, that server held the only copy of the code.

STEP_02 · Back up

A full copy, off the server

Panel backups plus everything they skip, pulled to a separate machine and verified with checksums.

STEP_03 · Build

A matching new server

HestiaCP with the same web, mail and database components, plus PostgreSQL, Node.js and PM2. I worked with the cloud provider to open the mail and web ports at their firewall.

STEP_04 · Restore

Everything running, privately

Panel users, mailboxes, DKIM keys and certificates restored; apps started under PM2 with the same names and ports; sites tested before any visitor could reach them.

STEP_05 · Switch

A short, planned cutover

DNS lifetimes lowered to 5 minutes days ahead, a content freeze, a final sync, then the switch. The old server was stopped so nothing could land in the wrong place.

STEP_06 · Prove it

Every email accounted for

Each database table, mailbox and API response compared with the old server. All 3,683 emails found on the new one, matched by message ID.

[02] // EMAIL_THAT_ARRIVES

Company email that lands in the inbox, not in spam

Self-hosted mail only works if the big providers trust the server. I set up every signal Gmail and Outlook check, and gave Bach Saiss its own mail domain for the first time.

  • SPF·DKIM·DMARCPublished and checked on all three domains.
  • Reverse DNSThe server's IP now answers with its real name, set through the provider.
  • BouncesFixed a missing record that made Gmail reject automatic replies.
  • EncryptionLet's Encrypt certificates on mail and webmail, renewed automatically.
  • ReputationNot listed on Spamhaus, SpamCop, Barracuda or SORBS.
  • WebmailRoundcube on each domain, plus spam and virus filtering.
[03] // ONE_SERVER_SEPARATE_KEYS

Each business manages its own corner

One server keeps costs down, but each business should only see its own sites and mailboxes. I split them into separate HestiaCP accounts, each with its own branded panel address.

panel.bachsaiss.comBach Saiss managerIts site, CMS and mailboxes only
panel.leqta.comLeqta managerIts site, CMS and mailboxes only
panel.[third business]Third businessIts apps and mailboxes only
own login, own HTTPS certificate, no access to web server templates
one VPS · administrator viewHestiaCP
nginx + Apache · MariaDB · PostgreSQL · Exim · Dovecot · PM2
[Screenshot · HestiaCP panel, web domains list for one business (blur the addresses)]
[04] // AFTER_THE_MOVE

Regular check-ups, inside and out

I check the server from the outside like a visitor would, then from the inside: services, apps, certificates, mail queue, logs. A few things those check-ups caught and fixed:

securityPassword guessing on mailBots were slowly guessing mailbox logins. A new fail2ban rule now bans them after 10 tries a day; it blocked 4 addresses within minutes.
mailLost forwards after panel updatesA panel rebuild quietly dropped some forwarding rules. I traced it and changed the setup so future rebuilds can't break it again.
webHTTPS everywhereForced HTTPS on every site, including a custom CMS template that was missing the hook, and cleaned up a web server name conflict.
upkeepUpdates with a way backSystem and kernel updates applied after a configuration backup, so any change can be undone.
[05] // ALSO_ON_MY_WATCH

A hardened VPS for app deployments

For OSCAR's tools I set up a separate Ubuntu VPS running Coolify, so apps deploy from Docker images with HTTPS handled automatically.

SSH hardenedfail2ban + repeat-offender bansfirewall: 22, 80, 443 onlyswap tunedCoolify · DockerLet's Encrypt
Next project · 03Bach Saiss