Closed GlennRicaud closed 4 years ago
I rebranded it to a bug. There are certain issues with quickfix.
"Search Request Scroll" is still not the best approach in cases when data must be temporary stored in memory (for extra processing - like in Vacuum). For huge datasets (like 5M versions in one repository) it either consumes entire memory or risks expiring cursor in elasticsearch.
Next best candidate algorithm is batching with sort by unique value and filtering of already processed entries by range filter.