Open paul-butcher opened 1 month ago
Actually, what about a Step Function? Something like this:
The Restore function could run "tonight" and emit the list of shoot numbers with the date "tomorrow", broken up into three groups spread across the day.
Actually. All of this is too complicated. We can have two lambdas, one that runs daily at 1930, another running daily at 0830
If this turns out to overwhelm Archivematica, then we can interpose a step where Restore batches them into 20s and a separate lambda pulls a single batch of 20 three times a day to put them all on the transfer queue. But we should cross that bridge when we come to it, because (🤞) AM will be a bit more reliable next time.
I've just had an experience of 60 at a time killing Archivematica, by filling up the failed folder, but that could have been due to some records failing yesterday and clogging up the place.
In order to get closer to a fire-and-forget approach, write a script that:
See: https://wellcome.slack.com/archives/C02ANCYL90E/p1723028197261449