A managed database, end to end

• DB Starter

  Shared-tier Postgres · 1 GB RAM · 10 GB storage

  Side projects, small apps

  Daily backups · 7-day retention

  $12/mo

• Recommended

  DB Standard

  Postgres · 2 GB RAM · 40 GB storage

  Production apps

  7-day PITR

  $25/mo

• DB Plus

  Postgres · 4 GB RAM · 100 GB storage

  Heavier production workloads

  30-day PITR

  $49/mo

Full rate detail, add-ons and fair-use limits live on the rate card →Pricing

• Records with a fixed shape

  Users, orders, subscriptions, comments.

  Rows an app reads and writes constantly. They must stay consistent with each other. This is most of what a database is for.

• Anything that must survive a restart

  A shopping cart, a login session, a saved draft.

  If losing it on a redeploy would be a real problem, it belongs here, not held only in a running process.

• What it is not for

  An uploaded photo, a generated report file.

  Those are files, not rows, and they live on a different kind of storage. See Files.

1. 1

   Your app holds one secret

   A single connection string, kept the way any other secret is — never written into the code itself. See Access.

   You

2. 2

   The path between app and database is private

   Not reachable from the open internet at all — only from the app it is attached to.

   CustomHosted

3. 3

   A connection comes from a shared pool, not one at a time

   A pool is a small set of open connections an app reuses, instead of opening and closing one for every query. It is faster. It is also what has a limit.

   CustomHosted

4. 4

   The tier sets that limit

   Every tier caps how many connections it will hold open at once. Staying near that cap for more than a spike is the outgrown signal on the ladder below. See Scaling.

   CustomHosted

1. 1

   The migration is written

   A new column, a new index, a changed constraint — the migration is application code. Write it yourself, or let a CodeHerder brief open it as a pull request.

   You

2. 2

   Tried against a copy first

   Pulled in on request from a recent backup, or seeded with placeholder data if a real copy isn't needed for the check being done. See Environments.

   CustomHosted

3. 3

   A snapshot is taken before it runs

   Nightly dump, plus point-in-time recovery on the tiers that carry it. See Recovery.

   CustomHosted

4. 4

   It runs inside the same deploy

   No separate migration step to remember. The pipeline that ships the code runs the migration too.

   CustomHosted

5. 5

   If it fails halfway

   The pipeline halts and the last known-good version keeps serving; a human is notified. See Monitoring.

   CustomHosted

You do not have to write the migration yourself. Describe the change in plain language instead. From the brief onward it takes the same path as every other reviewed change on this platform, which Patching describes in full. That is the Agentic iteration — CodeHerder add-on ($25/mo):

1. 1

   Briefed as a change

   Written up the same way any other change request is — this is what the Agentic iteration — CodeHerder add-on covers.

   CustomLabs

2. 2

   An agent opens the pull request

   The upgrade lands as a real pull request against the repository, with the test suite run against it.

   CodeHerder

3. 3

   A human reviews and merges

   Someone reads the diff before it goes anywhere near production — nothing here is auto-applied blind.

   CustomLabs

4. 4

   It ships through the same pipeline

   The same pipeline that deploys everything else on your stack ships this too — no separate process to babysit.

   CustomHosted

5. 5

   You're told it's done

   A note confirming what changed and that it merged, in the same shape as every other note on this site.

   CustomHosted

• ✓Automated daily backups
• ✓Point-in-time recovery
• ✓Managed patching & minor-version upgrades
• ✓Monitoring & alerts
• ✓Encrypted at rest and in transit
• ✓A human who answers

The tick list above promises minor-version upgrades. Version upgrades run on two tracks. What changes between them is who asks for the move, and when it runs:

• Minor version

  Standing window, monthly

  A brief connection blip, noted after the fact. See Patching.

  CustomHosted

• Major version

  Only when you ask, in a window you pick

  A major release can change how the engine behaves, so it never applies itself. You ask for the move. It runs the same way as a tier change: scheduled, snapshotted first, rollback ready. See Scaling.

  CustomHosted

A retention window is only useful if a restore from it actually works — see what a restore actually involves.

1. 1

   Auto-handled

   Restart, retry, or redeploy the last known-good version, done automatically

   No one, unless it recurs — it goes in the log, not a message to you

2. 2

   You hear from us

   A human looks at it and decides what to do — revert, restore, or a fix in code

   You, in writing, on the same quote channel as everything else — business hours for anything that isn’t urgent

3. 3

   Outage handling

   Worked ahead of the routine queue

   You, as soon as there is something to say, with priority handling; out-of-hours response is best effort, not a staffed round-the-clock desk

"When a bigger box is the wrong answer" is on Scaling.

• A one-time import, on the way in

  A dump from wherever the database runs today, restored into the new tier ahead of the cutover. Checked before anything points at it for real. See Migrations.

• A standard dump, on the way out

  Standard dump format. Restore into any compatible database. See Portability.

• Storage is billed inside the tier, not on the instance disk

  Sized and billed as part of the tier itself: 10 to 100 GB, depending which one. That is a different disk from the one an app writes to. See Files.

• A longer backup window is the one add-on

  Extended backup retention (30-day), +$8/mo — daily backups kept for 30 days instead of 7.

• This site keeps no database at all

  The one form on customhosted.com posts to a relay and keeps nothing — no row is ever written or read back.

• That is the right call for a marketing site

  A quote form does not need to survive a restart. It does not need a later query either. Adding a database here would add risk, not value.

• The moment that stops being true

  The day this site needs to show a customer their own account or history, it needs a real database. One of the tiers on this page would do — sized the same way as any other app.

• No read replicas today

  When read traffic alone is the bottleneck, a second database built for reads is a rebuild, not a resize. See Scaling for the rest of it.

• No automatic failover to a second machine

  One instance is one instance. Losing it triggers a restore, the same procedure as any other data-loss event. There is no silent switch to a standby. See Recovery for the rest of it.

• Query tuning is consulting work, not a managed feature

  A slow query needs a person to read it and rewrite it. That is real engineering, not a setting anyone flips. It is CustomLabs work, quoted like any other change.

• Postgres today, another engine on request

  This page prices Postgres. A different engine is a custom quote, sized and reviewed the same way as any other bespoke shape. See Custom for the rest of it.

1. 01What actually happens if too many connections open at once?

   Each tier caps how many it will hold open. Running near that cap for more than a brief spike is exactly the "outgrown" signal on the ladder above — it gets flagged, and the fix is either a bigger tier or fixing a connection leak in the app.

2. 02Who writes the schema migration?

   You do, or a CodeHerder brief does on your behalf as a pull request. Everything after that — the snapshot, running it inside the deploy, what happens if it fails halfway — is ours.

3. 03What happens if a migration fails halfway through?

   The same thing that happens to any other deploy that fails partway: the pipeline halts and the last known-good version keeps serving, with a human notified. The pre-migration snapshot is there if anything needs restoring from before it ran.

4. 04Who upgrades the database version?

   Minor versions move on the same standing patch window as everything else — a brief connection blip, noted after the fact. A major version never applies itself; you ask for it, and it runs on a window you pick, the same way a tier change does.

5. 05Do you offer read replicas or automatic failover?

   Not today. A read replica is a rebuild CustomLabs can design once read traffic alone is the bottleneck, not a setting to switch on. Losing the one instance is a restore, the same procedure as any other data-loss event.

6. 06Can you help with a slow query?

   Query tuning is real engineering, not a managed feature we flip on — it's CustomLabs work, designed and quoted like any other change.

7. 07Does this database ever hold my customers' data?

   This site's own form doesn't — it posts to a relay and keeps nothing. The tiers on this page exist for your app, sized the same way whether it holds ten rows or ten million.