Repository navigation
Conversation
SOP 节点产出的正文按字节存进 run_artifacts.content,而该列是 TEXT(MySQL 上限 65,535 字节),引擎与执行机器却允许单产物到 1 MiB(maxArtifactBytes / maxArtifactUploadBytes)。一份 74,768 字节的 design.md 就让整次 run 写入报 Error 1406 (Data too long for column 'content')。 更麻烦的是失败没人处理:节点终态走 updateRun,只把错误打日志,于是节点在库里 一直是 running、lease 一直续,run 永远不会结束也不会报错。 - run_artifacts.content 改为 MEDIUMTEXT,覆盖引擎 1 MiB 上限;附 migrations/034 幂等脚本给已有库升级(新建库由 AutoMigrate 直接建对类型)。 - 节点终态改用 updateRunAtErr:写入失败时先记一条 level=error 遥测,再把节点 置为 failed 并带上真实原因,而不是静默挂死。 - 新增 MySQL 回归测试:先还原 TEXT 列复现 1406,再回放 034 脚本验证 74KB 正文 能完整落库。
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
现象
一个 script/agent 节点产出 74,768 字节的
design.md后,run 永远停在design = running:之后没有任何进展,但 run 的 lease 一直在续(引擎认为它在跑),页面看起来像"还在执行"。
根因
run_artifacts.content,而该列是 MySQLTEXT(65,535 字节)。引擎侧maxArtifactBytes(internal/engine/artifacts.go)和执行机器侧maxArtifactUploadBytes(internal/runner/daemon.go)都允许单产物到 1 MiB,两个上限不一致;几千行的设计文档 / 报告很容易越过 64 KB。e.updateRun,写库失败只log.Printf(internal/engine/run_state.go),节点状态没落库,也没有任何东西会重调度它 —— 于是永久 hanging,而不是报错。改动
run_artifacts.content→MEDIUMTEXT(16 MiB),覆盖引擎的 1 MiB 上限;新建库由 AutoMigrate 直接建对类型。migrations/034_run_artifacts_content_mediumtext.sql给已有库升级(按仓库约定数据/DDL 修复走 migrations,手动执行)。internal/engine/node_execute.go:节点成功终态改用updateRunAtErr;写入失败时记一条level=error的node.commit.failed遥测,并把节点置为failed带上真实原因,避免静默挂死。committed=false(run 已不在 / lease 归别的实例)仍按原语义跳过。internal/repo/run_artifact_content_mysql_test.go:先把列还原成TEXT复现 1406,再回放 034 脚本,验证 74,768 字节正文能完整写回。(与其他 MySQL 专用测试一样由AGENTFLOW_RUN_ARTIFACT_TEST_DSN环境变量开启,默认 skip。)验证
GOOS=linux go build ./internal/engine/ ./internal/repo/ ./internal/runner/通过,go vet通过。TestArtifactsAreVersionedAndMaterializedForDownstreamNode、TestAutomaticGatesAndArtifactsCompleteBeforeManualApproval、TestShellAndPythonScriptNodesExecuteInOrder、TestArtifactCollectionErrorsAndFreeze全部 PASS(需要AGENTFLOW_GATE_MCP_BINARY指向 gate-mcp,属既有环境依赖)。run_artifacts.content改成mediumtext后,同一份 74 KB 产物可以正常落库,后续节点继续执行。备注
这次也顺手确认了 runner 侧
collectRunnerInvocationArtifacts对超过 1 MiB 的产物是静默丢弃(len(data) > maxArtifactUploadBytes直接 continue),行为与"产物 0 个"一致;那条路径本次没有产生问题,先不动,如果需要可以单独提。