This PR introduces a new `enable_experimental_backup_uploads` site setting (default false and hidden), which when enabled alongside `enable_direct_s3_uploads` will allow for direct S3 multipart uploads of backup .tar.gz files.
To make multipart external uploads work with both the S3BackupStore and the S3Store, I've had to move several methods out of S3Store and into S3Helper, including:
* presigned_url
* create_multipart
* abort_multipart
* complete_multipart
* presign_multipart_part
* list_multipart_parts
Then, S3Store and S3BackupStore either delegate directly to S3Helper or have their own special methods to call S3Helper for these methods. FileStore.temporary_upload_path has also removed its dependence on upload_path, and can now be used interchangeably between the stores. A similar change was made in the frontend as well, moving the multipart related JS code out of ComposerUppyUpload and into a mixin of its own, so it can also be used by UppyUploadMixin.
Some changes to ExternalUploadManager had to be made here as well. The backup direct uploads do not need an Upload record made for them in the database, so they can be moved to their final S3 resting place when completing the multipart upload.
This changeset is not perfect; it introduces some special cases in UploadController to handle backups that was previously in BackupController, because UploadController is where the multipart routes are located. A subsequent pull request will pull these routes into a module or some other sharing pattern, along with hooks, so the backup controller and the upload controller (and any future controllers that may need them) can include these routes in a nicer way.
This commit introduces a new s3:ensure_cors_rules rake task
that is run as a prerequisite to s3:upload_assets. This rake
task calls out to the S3CorsRulesets class to ensure that
the 3 relevant sets of CORS rules are applied, depending on
site settings:
* assets
* direct S3 backups
* direct S3 uploads
This works for both Global S3 settings and Database S3 settings
(the latter set directly via SiteSetting).
As it is, only one rule can be applied, which is generally
the assets rule as it is called first. This commit changes
the ensure_cors! method to be able to apply new rules as
well as the existing ones.
This commit also slightly changes the existing rules to cover
direct S3 uploads via uppy, especially multipart, which requires
some more headers.
In preparation for adding automatic CORS rules creation
for direct S3 uploads, I am adding tests here and moving the
CORS rule definitions into a dedicated class so they are all
in the one place.
There is a problem with ensure_cors! as well -- if there is
already a CORS rule defined (presumably the asset one) then
we do nothing and do not apply the new rule. This means that
the S3BackupStore.ensure_cors method does nothing right now
if the assets rule is already defined, and it will mean the
same for any direct S3 upload rules I add for uppy. We need
to be able to add more rules, not just one.
This is not a problem on our hosting because we define the
rules at an infra level.
Discourse automatically sends a private message after backup or
restore finished. The private message used to contain the log inline
even when it was very long. A very long log can create issues because
the length of the post will be over the maximum allowed length of a
post. When that happens, Discourse will try to create an upload with
the logs. If that fails, it will trim the log and inline it.
You can use `discourse restore --location=local FILENAME` if you want to restore a backup that is stored locally even though the `backup_location` has the value `s3`.
It pauses Sidekiq, clears Redis (namespaced to the current site), clears Sidekiq jobs for the current site, restores the database and unpauses Sidekiq. Previously it stayed paused until the end of the restore.
Redis is cleared because we don't want any old data lying around (e.g. old Sidekiq jobs). Most data in Redis is prefixed with the name of the multisite, but Sidekiq jobs in a multisite are all stored in the same keys. So, deleting those jobs requires a little bit more logic.
After restoring a backup it takes up to 48 hours for uploads stored on S3 to appear in the S3 inventory. This change prevents alerts about missing uploads by preventing the EnsureS3UploadsExistence job from running in the first 48 hours after a restore. During the restore it deletes the count of missing uploads from the PluginStore, so that an alert isn't triggered by an old number.
This is a follow-up for f51ccea028 because specs were failing on macOS. macOS uses bsdtar by default and the --transform option isn't supported by bsdtar. But the -s option is and it works the same.
This also ensures that restoring a backup works when it was created with the wrong upload paths in the time between ab4c0a4970 (shortly after v2.6.0.beta1) and this fix.
The `EXECUTE FUNCTION` syntax for `CREATE TRIGGER` statements was introduced in PostgreSQL 11. We need to replace `EXECUTE FUNCTION` with `EXECUTE PROCEDURE` in order to be able to restore backups created with PG12 on PG10.
* DEV: new S3 backup layout
Currently, with $S3_BACKUP_BUCKET of "bucket/backups", multisite backups
end up in "bucket/backups/backups/dbname/" and single-site will be in
"bucket/backups/".
Both _should_ be in "bucket/backups/dbname/"
- remove MULTISITE_PREFIX,
- always include dbname,
- method to move to the new prefix
- job to call the method
* SPEC: add tests for `VacateLegacyPrefixBackups` onceoff job.
Co-authored-by: Vinoth Kannan <vinothkannan@vinkas.com>
This reverts commit 20780a1eee.
* SECURITY: re-adds accidentally reverted commit:
03d26cd6: ensure embed_url contains valid http(s) uri
* when the merge commit e62a85cf was reverted, git chose the 2660c2e2 parent to land on
instead of the 03d26cd6 parent (which contains security fixes)
Rails calls I18n.translate during initialization and by default translation overrides are used. Database migrations would fail if the system tried to migrate from an old version that didn't have the `translation_overrides` table with all its columns yet.
This makes restoring really old backups work again. Running `DISABLE_TRANSLATION_OVERRIDES=1 rake db:migrate` will allow you to upgrade such an old database as well.
Some specs use psql to test database restores and dropping the table after the test needs to happen outside of rspec because of transactions. The previous attempt lead to some changes to be stored in the test database.
That commit introduced a bug to the system: f69dacf979
Restore works fine for multisite, however, stopped working for non-multisite.
Reason for that was that `establish_connection` method got a check if the multisite instance is available:
```
def self.instance
@instance
end
def self.establish_connection(opts)
@instance.establish_connection(opts) if @instance
end
```
However, the reload method don't have that check
```
def self.reload
@instance = new(instance.config_filename)
end
```
To solve it, let's ensure we are in a multisite environment before call reload
Zeitwerk simplifies working with dependencies in dev and makes it easier reloading class chains.
We no longer need to use Rails "require_dependency" anywhere and instead can just use standard
Ruby patterns to require files.
This is a far reaching change and we expect some followups here.
