Skip to content

New Search Workload Node Pool

Motivation

New Search displayed the system AKS node pool as the compute environment because the AKS monitor summary used the first agent pool. ElasticBLAST jobs run on the user workload pool, normally blastpool, so the form could show and submit the wrong SKU and node count.

User-facing change

The Compute Environment section now displays the workload node pool, preferring blastpool, then any mode=User pool. The submit payload uses the same workload pool for machine_type and num_nodes.

API/IaC diff summary

  • api.services.monitoring.list_aks_clusters now summarizes AKS node_sku and node_count from the workload pool while preserving the full agent_pools list.
  • New Search frontend derives its displayed pool, SKU, node count, sharding preview, and submit payload from the workload pool.
  • No IaC changes.

Validation evidence

  • Added api/tests/test_monitoring_aks_pools.py for blastpool and generic user-pool fallback selection.
  • uv run pytest -q api/tests/test_monitoring_aks_pools.py -> 2 passed.
  • uv run pytest -q api/tests -> 265 passed.
  • cd web && npm run build -> built successfully.