Skip to content
PostgreSQL PITR restore fails with ENOSPC on Hobby 5 GB volume
ahmadawwwad
OP

4 hours ago

Subject: PostgreSQL PITR restore fails on 5 GB volume

Hello Railway Support,

I'm experiencing a PostgreSQL Point-in-Time Recovery issue on my LAMMA project.

PITR is enabled and healthy, and my production database uses approximately 350 MB.

I attempted two restores into separate services:

  • November 18, 2026, at 08:00 UTC.
  • November 18, 2026, at 07:00 UTC.

Both attempts failed with:

No space left on device

The restored volumes reached approximately 5 GB during WAL recovery, and PostgreSQL never became ready to accept connections.

The original production database is healthy and unaffected. I have deleted both failed restore services.

Could you please advise how to reliably restore this database within the Hobby plan's 5 GB volume limit, without upgrading?

Thank you.

Awaiting User Response

1 Reply

Railway
BOT

4 hours ago

A point-in-time restore has to hold the base backup plus all of the write-ahead log it replays between that backup and your chosen target, so the space it needs depends on how much your database wrote in that window, not on its 350 MB size. Both restored volumes were already created at 5 GB, the largest volume Hobby allows, so the replay outgrew the maximum size.

On Hobby, the way to make the restore fit is to pick a restore target closer to a recent backup, so there is less write-ahead log to replay. If the point you need is far enough from a backup that the replay needs more than 5 GB, a plan with a higher volume limit is the only way to give the restore more room.

A failed restore does not touch the source database, and the two leftover restore volumes are already scheduled for deletion.


Status changed to Awaiting User Response Railway • about 4 hours ago


Welcome!

Sign in to your Railway account to join the conversation.

Loading...