Apparently some instances have local users with local ap_ids
that are marked as local: false. Needs more investigation into how
that happened.
In the meantime, the skeleton migration was patched to just ignore
all known ap ids, not just locals. Doesn't seem to slow down the migration
too much on patch.cx
Closes #1746
Repo,
"select distinct unnest(nonexistent_locals.recipients) from activities, lateral (select array_agg(recipient) as recipients from unnest(activities.recipients) as recipient where recipient similar to '#{
instance_uri
- }/users/[A-Za-z0-9]*' and not(recipient in (select ap_id from users where local = true))) nonexistent_locals;",
+ }/users/[A-Za-z0-9]*' and not(recipient in (select ap_id from users))) nonexistent_locals;",
[],
timeout: :infinity
)