perf: streamline repository and deployment awareness
ForgeFlow quality gate / quality (push) Canceled after 0s
ForgeFlow quality gate / quality (push) Canceled after 0s
This commit is contained in:
@@ -0,0 +1,12 @@
|
||||
# ForgeFlow 0.10.12
|
||||
|
||||
## Faster awareness with stricter deployment truth
|
||||
|
||||
- Repository refreshes are **coalesced** and briefly cache Gitea inventory and workspace discovery; a manual refresh remains fully forced and file changes arriving mid-refresh receive one trailing refresh.
|
||||
- Server discovery reuses its Docker and Compose evidence for existing deployment profiles instead of opening a separate SSH session for every linked workload.
|
||||
- Deployment status only claims **exact Gitea commit parity** after comparing a concrete branch SHA with the live server SHA; matching repository provenance alone is no longer sufficient.
|
||||
- Container discovery uses **batched Docker inspect** with a safe per-container fallback when a container disappears during the scan.
|
||||
- Active Gitea and SSH deployment polling uses **bounded worker pools**, improving multi-deployment latency without flooding external services.
|
||||
- A **stopped container** can no longer be marked healthy because another process answers on its previous healthcheck port.
|
||||
- Large repository and server-inventory lists use offscreen rendering containment to reduce layout and paint work.
|
||||
- Inventory diagnostics now include scan and state-refresh durations, and Gitea bulk verification fails fast after a confirmed connectivity outage.
|
||||
Reference in New Issue
Block a user