Merge pull request #10310 from demarches-simplifiees/fix_cron_purge_unattached_blobs

Tech: debloque un le cron de nettoyage des blobs en limitant le périmètre à une semaine
This commit is contained in:
LeSim 2024-04-15 08:56:53 +00:00 committed by GitHub
commit 224f755a51
No known key found for this signature in database
GPG key ID: B5690EEEBB952194

View file

@ -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