You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
This repository was archived by the owner on Nov 21, 2026. It is now read-only.
Repository navigation
This repository was archived by the owner on Nov 21, 2026. It is now read-only.
benchmarks: a default-off override for ArcadeDB's queryMaxHeapRAM, stamped on the rows, for the tpch10 olap repair stage (DECISIONS #175) #260
ArcadeDB 26.10.1's query heap budget (default half the heap) refuses TPC-H SF10 top_parts (about 2M groups, 12 cores, 16 GB heap) in both ArcadeDB arms of the l1tpc lane, where the 2026-09-24 snapshot ran it (ArcadeData#9402; DECISIONS #175). Both options for the headline need rows measured with the budget raised, so this adds the mechanism, default OFF so the running chain is unchanged: an environment variable BENCH_ARCADE_QUERY_MAX_HEAP_RAM_MB that adds -Darcadedb.queryMaxHeapRAM=N to the embedded JVM (l1_tpc) and to the served arm's JAVA_OPTS, a row field arcadedb_query_max_heap_ram_mb with a source (read back from the engine where it can be asked, the request otherwise), and an overrides.py entry so the generated sentence and the gates know the setting when rows carry it. The repair stage itself is generated after qP09.
Activity
benchmarks: a default-off, stamped override for ArcadeDB's query heap…