You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
When importing backups, files that are stored on R2 are inline linked as if S3 isn't enabled for some reason, rebaking the posts doesn't change anything.
Affected files are most optimized profile pics, site logos and new files uploaded since enabling S3.
Notable: migrate_to_s3 task also doesn't run correctly.
Even if I wanted to include S3 uploads into the backups, there is something missing in the R2-S3 compatible API stopping discourse from downloading the images in the backup process.
I can currently only move the server using by just copying the bind volumes, which isn't great and isn't how discourse backups work.
The plan
write a post about the S3 content link issue on meta
Maybe there's something that can fix this, because the files actually exist on R2, but the new discourse instance isn't using the R2 CDN-link for some reason.
Move off S3
the only officially supported S3 storage solution is AWS, to avoid further issues, we either move to AWS S3, which is very unpredictable in terms of monthly costs, or we disable S3 in discourse completely.
Because Discourse can't currently download uploads from R2 to include in the backups, we first need to move to a fully S3 compliant solution.
Cloning the bucket to a local minio instance, and then letting discourse use that as the new S3 host seems to be the best workaround.
The text was updated successfully, but these errors were encountered:
Discourse backups don't work.
When importing backups, files that are stored on R2 are inline linked as if S3 isn't enabled for some reason, rebaking the posts doesn't change anything.
Affected files are most optimized profile pics, site logos and new files uploaded since enabling S3.
Notable: migrate_to_s3 task also doesn't run correctly.
Even if I wanted to include S3 uploads into the backups, there is something missing in the R2-S3 compatible API stopping discourse from downloading the images in the backup process.
I can currently only move the server using by just copying the bind volumes, which isn't great and isn't how discourse backups work.
The plan
write a post about the S3 content link issue on meta
Maybe there's something that can fix this, because the files actually exist on R2, but the new discourse instance isn't using the R2 CDN-link for some reason.
Move off S3
the only officially supported S3 storage solution is AWS, to avoid further issues, we either move to AWS S3, which is very unpredictable in terms of monthly costs, or we disable S3 in discourse completely.
Because Discourse can't currently download uploads from R2 to include in the backups, we first need to move to a fully S3 compliant solution.
Cloning the bucket to a local minio instance, and then letting discourse use that as the new S3 host seems to be the best workaround.
The text was updated successfully, but these errors were encountered: