Prefect crashes my services #18193
Replies: 2 comments 2 replies
-
hi @slavaGanzin - thanks for sharing your thoughts! I converted this to a discussion because there wasn't a specific request here. I understand your concerns with respect to "interrupting real work with bookkeeping" etc but its most constructive if issues remain well-scoped to concrete bugs or enhancements. Discussions are great for getting your thoughts out and potentially catalyzing more specific issues. In general we do attempt to avoid needless disruption of user code, so if you have specific examples where this is not the case (preferably with MREs!), we'd be happy to see them! additionally, the error you encountered with |
Beta Was this translation helpful? Give feedback.
-
Again, I don't care about prefect being unable to parse my logs, why should my logs contain this? --- Logging error --- |
Beta Was this translation helpful? Give feedback.
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
-
Bug summary
Hello, I like prefect very much and it's a huge step up from airflow-like garbage.
But there is a problem with python "let it fall" approach, because in contrast with Erlang were this idea emerged, Python doesn't work on the actors and messages concept because it is tragically synchronous (hello GIL and "asynchronous" asyncio library).
And also there is a problem with the fact that prefect is observability platform, it observes if something is working or not and that's it. You may think that by wrapping it into flow and task, you became a task workflow orchestration mumbo jumbo, but it's not: my code run some tasks, my code do some job, prefect - observe.
So in general, I don't care that much if prefect "Couln't authorize", "You limit is reached" or any other problems with prefect, if my tasks continue working. But I found out, that prefect have such inflated ego, that it crashes my services, because it couldn't login. Are you serious?
And you know what happens? I cannot get alerts from my prefect because my service couldn't log in to indicate that it failed. So I need to periodically check manually if my jobs are working fine. Then I may ask, for what reason do I have an observability tool.
Maybe if one of my services couldn't login, you may show me that in prefect UI and not silently crash my service? Just a suggestion
You need to think about it. But here is a quick fix.
You are digging a hole under yourself, unreliable workflow orchestration is worse than having none. I recommended prefect numerous times as a go to substitution for airflow and others, and now I feel uneasy about it
Version info
Additional context
No response
Beta Was this translation helpful? Give feedback.
All reactions