Deploy to Production / pre-check (push) Successful in 54s
Deploy to Production / build (push) Successful in 9m8s
Deploy to Production / deploy-edgeone (push) Successful in 5m29s
Deploy to Production / deploy-upyun (push) Successful in 6s
Deploy to Production / finalize (push) Successful in 4m17s
Deploy to Production / notify-failure (push) Skipped
根因:/opt/gitea/runner/config.yaml 中 container.network=gitea_default (job 容器落在 172.20.x),但 runner 容器自身只挂在 bridge 网络 (172.17.0.2),其提供的缓存服务 172.17.0.2:44869 跨 bridge 网段不可达。 实测证据: - runner 容器内 wget http://127.0.0.1:44869/ -> HTTP 404(服务活着) - gitea 容器 wget http://172.17.0.2:44869/ -> download timed out - 构建日志:Setup Node restore 死等 278s(01:07:17 -> 01:11:55 ETIMEDOUT) Post Setup Node save 死等 288s(01:14:54 -> 01:19:42 ETIMEDOUT) Cache Hugo resources restore 亦超时 合计每次构建约 10 分钟纯浪费,且缓存从未真正命中或保存成功过。 改动: - Setup Node 的 cache 改为按 github.server_url 条件启用;Gitea 上 表达式求值为空串,setup-node 自动跳过缓存 - Cache Hugo resources 步骤加 if 条件,Gitea 上不执行(避免 restore+save 双超时) - npm ci 增加 --no-audit --no-fund - step summary 在 Gitea 上如实显示"已跳过",不再误报"未命中" 无副作用:npm ci 实测 68s 本就是在无缓存状态下跑的(restore 从未成功), 去掉缓存不会使其变慢;GitHub 侧行为完全不变。