Change Details

At least the metadata indexer should be, because currently, it is based on a journal client dedicated to creating one-shot tasks. This would simplify the stack by removing a moving part. Additionally, it would be an opportunity to make sure indexation is always retried on error (currently, it isn't).

At least the metadata indexer should be, because currently, it is based on a journal client dedicated to creating one-shot tasks. This would simplify the stack by removing a moving part. Additionally, it would be an opportunity to make sure indexation is always retried on error (currently, it isn't; and failing to index one object causes the whole batch to be lost).

At least the metadata indexer should be, because currently, it is based on a journal client dedicated to creating one-shot tasks. This would simplify the stack by removing a moving part. Additionally, it would be an opportunity to make sure indexation is always retried on error (currently, it isn't; and failing to index one object causes the whole batch to be lost).