Athena vocabulary load is too slow and unstable on large tables. The concept_ancestor table (35 M rows) and concept (6 M rows) could take up to 1h (depending on local resources) to load via the existing ORM-based path, and the entire merge happened in a single transaction. This caused unbounded WAL accumulation before the commit, which led to instability on memory-constrained hosts.
There is also no dedicated load-vocab-source command. Vocabulary loading was bundled into a general-purpose CLI without bulk-mode controls (disabling indexes and FK triggers around the load) or any progress feedback.
Athena vocabulary load is too slow and unstable on large tables. The
concept_ancestortable (35 M rows) andconcept(6 M rows) could take up to 1h (depending on local resources) to load via the existing ORM-based path, and the entire merge happened in a single transaction. This caused unbounded WAL accumulation before the commit, which led to instability on memory-constrained hosts.There is also no dedicated
load-vocab-sourcecommand. Vocabulary loading was bundled into a general-purpose CLI without bulk-mode controls (disabling indexes and FK triggers around the load) or any progress feedback.