Archiving the public data of a forum

I’d like to enquire about what options there are for the public archiving of a forum.

We have recently migrated from phpBB, and we previously had a script that would archive the public posts to a text file. Are there any options like that for Discourse? We have about 30 public categories that would need to be archived. The username/display name of each poster would need to be included, but not the email or IP addresses or any other private data. SQL so that the data could be processed like Stack Exchange’s SEDE would probably be most useful.

Alright, so I came up with this. Seems to work well to me, though I need to double check that I am filtering out all the hidden posts correctly. And some of the columns probably aren’t actually that useful.

This can be used from a cronjob, but the user will need to be added to the docker group. I added a step after the dump to scp the file to the archive server, but you’d could do pretty much anything else with it too.


# Archive public data from the forum

EXEC="docker exec -i -u postgres -w /var/www/discourse app"

# Run commands to construct the exported tables
$EXEC psql discourse << EOF

SET client_min_messages TO warning;

DROP TABLE IF EXISTS users_export;
CREATE TABLE users_export AS
  SELECT id, username, created_at, name
  FROM users
  WHERE staged = false;

DROP TABLE IF EXISTS categories_export;
CREATE TABLE categories_export AS
  SELECT id, name, slug, parent_category_id
  FROM categories 
  WHERE read_restricted = false;

DROP TABLE IF EXISTS topics_export;
CREATE TABLE topics_export AS
  SELECT, t.title, t.last_posted_at, t.created_at, t.updated_at,
    t.views, t.posts_count, t.user_id, t.last_post_user_id, t.reply_count,
    t.category_id, t.closed, t.slug
  FROM topics AS t
  INNER JOIN categories_export AS c ON t.category_id =
  WHERE t.deleted_at is NULL AND t.archived = false;

DROP TABLE IF EXISTS posts_export;
CREATE TABLE posts_export AS
  SELECT, p.user_id, p.topic_id, p.raw, p.created_at, p.updated_at,
    p.reply_to_post_number, p.reply_count, p.quote_count, p.like_count
  FROM posts AS p
  INNER JOIN topics_export AS t ON p.topic_id =
  WHERE p.post_type = 1 AND p.deleted_at is NULL AND p.hidden = false;


# Dump data
echo Dumping data...
$EXEC pg_dump discourse -t users_export -t categories_export -t topics_export -t posts_export > forum-archive.sql

# Cleanup
$EXEC psql discourse << EOF

SET client_min_messages TO warning;

DROP TABLE IF EXISTS users_export;
DROP TABLE IF EXISTS categories_export;
DROP TABLE IF EXISTS topics_export;
DROP TABLE IF EXISTS posts_export;


This is something I am interested in, there are a few topics touching on this already.

You can get fairly close by crawling with a webcrawler user agent, and saving the results as static HTML files. There are tools that do this, as covered in some existing topics.

Might be something we could run an external project for @sam to make a “pretty” version of that.


Is the plan to put the site in “cold public storage” for a time and perhaps re-start the forum later? If that’s the case, then just saving the backup (rather than wanting to re-import the data back to Discourse later) solves that problem and would allow people to be re-connected with their accounts.

Or, do you want to periodically push off part of the live forum to another static site?

You want just the topics and not user pages? Are internal links important? Is a single text file per topic what you’re looking for, or would you prefer HTML that preserved at least some formatting and images and such?


It can’t be a full normal backup because that has lots of private information.

Previously, before we migrated, the archiving was partly for data preservation purposes. But now we have both Linode and S3 backups, so that’s not really necessary. Now an archive that would let you explore and process the data would be useful. If you’re familiar with Stack Exchange’s SEDE, that’s the sort of thing I’d like to enable. So a sanitised SQL dump would be best I think.

I’m going to install the data explorer plugin, then have a look at what columns there actually are. I might be able to figure out the sql commands I need myself.

Sorry to use you all as programming rubber ducks :wink: