diff --git a/app/jobs/cron/purge_unattached_blobs_job.rb b/app/jobs/cron/purge_unattached_blobs_job.rb index 9d04391cf..218b6f5d2 100644 --- a/app/jobs/cron/purge_unattached_blobs_job.rb +++ b/app/jobs/cron/purge_unattached_blobs_job.rb @@ -5,12 +5,19 @@ class Cron::PurgeUnattachedBlobsJob < Cron::CronJob # .in_batches { _1.each... } is more efficient in this case that in_batches.each_record or find_each # because it plucks only ids in a preliminary query, then load records with selected columns in batches by ids. # This is faster than other batch strategies, which load at once selected columns with an ORDER BY in the same query, triggering timeouts. - ActiveStorage::Blob.unattached.select(:id, :service_name, :created_at).in_batches do |relation| - relation.each do |blob| - return if blob.created_at > 24.hours.ago # not in where() because it's not an indexed column - - blob.purge_later - end + # + # .where(created_at: 1.week.ago..1.day.ago) to limit the number of records to be joined + # to the attachments table because of the unattached scope. Otherwise, it is triggering timeouts. + # + # the creation of an index on created_at does not seem required yet. + # + # caveats: the job needs to be run at least once a week to avoid missing blobs + ActiveStorage::Blob + .where(created_at: 1.week.ago..1.day.ago) + .unattached + .select(:id, :service_name) + .in_batches do |relation| + relation.each(&:purge_later) end end end