hydra: reduce eval worker memory to 4G
We want to see if we can get memory usage to spike less, to evaluate if we can colocate the database again.
Martin Weinelt committed
Jul 12, 2026 at 14:06 UTC
6a6ed0496f7b605481561d6c70578c02449af9f9
1 file changed
+1
-1
build/hydra.nix
+1
-1
@@ -94,7 +94,7 @@ in
94
log_prefix = https://cache.nixos.org/
95
96
evaluator_workers = 16
97
- evaluator_max_memory_size = 8192
97
+ evaluator_max_memory_size = 4096
98
99
queue_runner_endpoint = http://${config.services.hydra-queue-runner-dev.rest.address}:${toString config.services.hydra-queue-runner-dev.rest.port}
100