35 lines
1.5 KiB
Ruby
35 lines
1.5 KiB
Ruby
|
class ApplicationMailerJob < ActionMailer::MailDeliveryJob
|
||
|
|
# TODO: We have a mix of mailers that take ActiveRecords as arguments, and
|
||
|
|
# mailers that take IDs as arguments. If an item is unavailable when the
|
||
|
|
# notification is sent, it'll produce an ActiveJob::DeserializationError in
|
||
|
|
# the former case, and an ActiveRecord::RecordNotFound error in the latter.
|
||
|
|
#
|
||
|
|
# Ideally, we don't want to catch RecordNotFound errors, because they might
|
||
|
|
# be a sign of a different problem. But until we move all of the mailers over
|
||
|
|
# to taking ActiveRecords as arguments, we need to catch both errors.
|
||
|
|
|
||
|
|
retry_on ActiveJob::DeserializationError,
|
||
|
|
attempts: 3,
|
||
|
|
wait: 1.minute do
|
||
|
|
# silently discard job after 3 failures
|
||
|
|
end
|
||
|
|
|
||
|
|
retry_on ActiveRecord::RecordNotFound,
|
||
|
|
attempts: 3,
|
||
|
|
wait: 1.minute do
|
||
|
|
# silently discard job after 3 failures
|
||
|
|
end
|
||
|
|
|
||
|
|
# Retry three times when calling a method on a nil association, and then fall
|
||
|
|
# back on the Resque failure queue if the error persists (so that we have a
|
||
|
|
# record of it). This should address the issues that arise when we
|
||
|
|
# successfully load an object, but can't load its associations because the
|
||
|
|
# original object has been deleted in the meantime.
|
||
|
|
#
|
||
|
|
# (Most errors for calling a method on a nil object will be NoMethodErrors,
|
||
|
|
# but any that occur within the template itself will be converted to
|
||
|
|
# ActionView::Template:Errors. So we handle both.)
|
||
|
|
retry_on NoMethodError, attempts: 3, wait: 1.minute
|
||
|
|
retry_on ActionView::Template::Error, attempts: 3, wait: 1.minute
|
||
|
|
end
|