Supabase project Unhealthy or stuck hibernated (28P01, 503, restart refused): get it back without losing data
Your Lovable, Bolt, or Cursor app is down because the Supabase project shows Unhealthy, will not wake from hibernation, and the SQL Editor says 28P01 password authentication failed. What it means, what not to click, and how to get support to recover it in place.
Free · no card
Paste your error. Get a free diagnosis by email.
Tell us what broke and we’ll email you a short diagnosis: the likely cause and what to try first. Usually within a day.
What you are seeing
- Every login and data load in the app fails at once.
/auth/v1/healthand/rest/v1/...return503or521, or time out, and you changed no code. - The Dashboard shows the database, PostgREST, Auth, Realtime, or Storage as Unhealthy, or still says healthy while nothing works.
- The SQL Editor or Table Editor fails with
FATAL: 28P01: password authentication failed for user "postgres"(orsupabase_admin), although you never changed the database password. - Restart project is refused with
Project is currently hibernated and will wake on next supported request, but no request wakes it. Pausing can also fail, for example withFailed to verify backup status before pausing. - It is a Free plan project that had little traffic for a day or two before it went down.
Why it happens
- In the week of November 8, 2026, several Free plan projects in different regions (eu-west-1, eu-central-1, ap-northeast-2, ap-southeast-1) were reported on Supabase's GitHub stuck after an automatic hibernation or a restore: the database never came back up, Auth crash-looped because it could not reach Postgres, and self-serve restart was blocked.
- When Postgres is not running, the SQL Editor's connection fails at the pooler, so the error you see is
28P01even though your password is fine. In that case the password is a symptom, not the cause, and resetting it will not help. - There is also a real stale-password case: Supabase's changelog says a physical restore or an unpause could bring back database credentials from the backup snapshot. Supabase says restores now reapply current credentials, but if you rotated the password recently and connect through the shared pooler, cached credentials can also briefly cause
28P01. - Once the project is in this state there is no SQL or API call you can make to force Postgres to start. Supabase's infrastructure team has to recover it, which is why the community replies on these threads all point to a support ticket.
How to fix it
1. Confirm it is the project, not your code
curl -s -o /dev/null -w '%{http_code}\n' \
https://YOUR_PROJECT.supabase.co/auth/v1/health -H 'apikey: YOUR_ANON_KEY'
curl -s -o /dev/null -w '%{http_code}\n' \
'https://YOUR_PROJECT.supabase.co/rest/v1/your_table?select=id&limit=1' \
-H 'apikey: YOUR_ANON_KEY'
# 200 = the project is fine, the bug is in your app
# 503 / 521 / timeout on both = the project itself is down
# then check status.supabase.com for an incident in your region
2. Try one wake-up, then wait
Open the project in the Dashboard (opening it can trigger a wake-up), send one or two normal API requests like the curl above, then wait 10 to 15 minutes before trying again. Do not loop requests or reset the password over and over: per Supabase's docs, repeated authentication failures can get your IP banned or open the shared pooler's circuit breaker, which adds a second problem on top of the first.
3. Read the 28P01 correctly
If you did not change the database password, treat 28P01 in the SQL Editor as "the database is not running", not as a wrong password. If you did rotate it recently, follow Supabase's guide below: test a direct connection (db.YOUR_PROJECT.supabase.co) with the new password before blaming the pooler, and rotate only once.
4. Do not make it worse
While the project is Unhealthy, do not delete it, transfer it, restore an old backup over it, or create a new project and point your app at an empty database. If you have no backup of your own, the data inside this project is the only copy, so the goal is recovery in place.
5. Open a support ticket support can act on
Subject: Free project <ref> stuck hibernated / Unhealthy since <time UTC>, please recover in place
Project ref: <ref> Region: <region> Plan: Free
Down since: <first failure, UTC>
Errors:
- REST/Auth: 503 (or 521) on /auth/v1/health and /rest/v1/
- SQL Editor: FATAL: 28P01: password authentication failed for user "postgres"
(password not changed)
- Restart: "Project is currently hibernated and will wake on next supported request"
Tried: opened dashboard, API requests, waited <n> minutes
Request: please start Postgres and recover the project in place,
preserving tables, auth users and Storage objects.
This is a production app with real users.
# Send it from the Dashboard support form (or [email protected])
# and keep it to one ticket; add the ticket ID to any forum post.
6. Tell your users
Put a short maintenance notice on the app (a banner or a static page) saying login and data are temporarily unavailable and nothing is lost. If your frontend shows a blank screen or a raw error when Supabase is down, have it catch the failed request and show that message instead.
7. Once it is back: take your own backup
# needs the Supabase CLI and your database connection string
# (Dashboard > Connect; keep it out of git)
supabase db dump --db-url "$DB_URL" -f roles.sql --role-only
supabase db dump --db-url "$DB_URL" -f schema.sql
supabase db dump --db-url "$DB_URL" -f data.sql --use-copy --data-only
# run it on a schedule (e.g. a weekly GitHub Action that uploads the files
# somewhere private) so the next outage is never a data-loss question
8. Decide if Free is still the right plan
If paying users depend on the app, the Free plan's pausing and hibernation are a business risk, not just an inconvenience. Upgrading the organization to Pro removes inactivity pausing and gives you dashboard backups. For a demo or side project, the keep-alive in our project-pausing guide (linked below) plus your own backups is usually enough.
How to know it is fixed
Both curl checks return 200, the Dashboard shows every service as Healthy, the SQL Editor runs select 1;, and a real login, data load, and file upload work in the deployed app. You have a dated dump of your roles, schema, and data stored outside Supabase. Sources: Supabase discussion #53149 (Free project stuck after hibernation, Nov 9, 2026; same failure in #53150 and #53152), supabase/supabase #53147 (Unhealthy, 28P01, restart refused; maintainer reply: contact support), Supabase changelog: stale database credentials after a restore, Supabase: FATAL password authentication failed, Supabase: 28P01 after password rotation, and Supabase CLI: db dump. Related: Supabase paused my free project and Free plan over quota.