I’ve got a standard install that’s trying to restore a database. It is failing on the migration. ALTER TABLE Migrating the database... EXCEPTION: rake db:migrate Failed to migrate database. rake aborted! StandardError: An error has occurred, this and all later migrations canceled: (StandardError) …

OK. I’m restoring on a production site. It’s migrating. Here’s a query that’s been running for 30 minutes: discourse=# SELECT pid, usename, state, query, query_start FROM pg_stat_activity WHERE state = 'active'; pid | usename | state | query …

-rw-r--r-- 1 discourse discourse 16G Feb 10 21:13 site-2026-02-10-174058-v20250507013646.sql -rw-r--r-- 1 discourse discourse 2.9G Feb 10 21:11 site-2026-02-10-174058-v20250507013646.sql.gz root@forum-data:/shared# # after delete root@forum-data:/shared# du -hs postgres_data/; df -h . 24G post…

After the database had restored, postgres_data was 24G. Something in the migration had postgres_data balloon to 91G before the disk filled and the restore failed. This is on a clean install with the Discourse versions shown above. This seems like it must be a bug, so I’m recategorizing. I’ve also …

Do you have any hints how a 24G database could expand to 90G during database migration?

not sure.. I would probably run ncdu or something and have a look when the situation happened.

Using du and df I watched the size of postgres_data as it grew from 25g to 90g and the space on the disk go to (near) zero before it failed. I guess i need to find a way to track what query is running the next time. I’ve seen it on at least two sites. One with an upgrade and one on a restore. Bot…

Bunch of migrations rewrite whole tables, which can temporarily use more space, and PostgreSQL doesn’t aggressively gives space back to the system too. Does that forum has AI with embeddings enabled ?

[afbeelding] Falco: Bunch of migrations rewrite whole tables, which can temporarily use more space, and PostgreSQL doesn’t aggressively gives space back to the system too. I was thinking something like that. . . [afbeelding] Falco: Does that forum has AI with embeddings enabled ? That…

If you want to go down the rabbit hole, you could restore it manually to a dev environment that is at the May 2025 commit, then go to current commit and run migrations one by one and print the space before and after each — obviously with a script :stuck_out_tongue: .

Restore fails due to disk space on migration because of 70M calendar events

Ondersteuning

sam (Sam Saffron) 11 februari 2026 om 23:44 4

I am putting this in support, this message is coming direct from the file system, if PG thinks it has no space, it probably has no space cause not enough is carved out for it.

Topic		Antwoorden	Weergaven
Database migration chokes on huge value of a "calendar-details" item in table "post_custom_fields" Bug	13	2084	7 december 2020
'bundle exec rake db:migrate' taking a long time due to calendar migration Self-hosting	33	917	5 december 2023
Problem with restore Discourse from backup (3.4, quite large DB) Support	18	305	14 januari 2025
Database issues migrating form old version 20150822141540 Self-hosting	14	1039	22 oktober 2022
Restore Failing - Check Free Disk Space Self-hosting server-resources	11	1678	19 januari 2021

Restore fails due to disk space on migration because of 70M calendar events

Gerelateerde topics