Open JuchangGit opened 3 months ago
I have 15 dags scheduled at schedule: 0 9 * I checked the yaml. it is correct. but some of the dags did run at 9'o clock. and some of those didn't run.
Could you please share the scheduler's log? Are the DAGs on NFS or local disk?
The dags are on nfs. We can move them to local directory. Does nfs has something to do about it?
schedule: "/30 " schedule: "/20 " 我的都是没有问题的
schedule: "/33 *" 就有问题了,这样始终会每30分钟执行一次,而不是33分钟
The file event watcher may not function properly when DAG files are stored on NFS, which can prevent the scheduler from automatically reloading updated DAG files. In such cases, you'd need to manually restart the scheduler process to ensure it picks up the latest DAG files and their scheduling configurations. We've already implemented a polling mechanism as a workaround for NFS, instead of relying on file events. However, we're not entirely sure if this solution covers all scenarios.
For a more robust solution in future versions, I'm considering reading DAGs directly from a GitHub branch or some type of database. This approach should resolve these issues.
我用的不是nfs,我的问题不是没有调度起来,而是当调用频次不是整 十分钟 的时候 调度会近似到 整十分钟
我可能没有说明白,看看我的截图
正确的调度表应该是 1:33 2:03 2:36 3:09 3:42 ...
我的 schedule: "/33 *" 但是执行时间表是: 00:00:00 00:33:00 01:00:00 01:33:00 ... 这并不是每33分钟执行一次