Hi, when I deploy my cluster (on AWS fargate) the distributed cluster seems to think it is the Unix Epoch instead of the real time. How can I fix this?
It actually started a little bit before the Epoch.
I did a manual deployment with my own Docker images.
Thank you!
Hi @Joseph-Rubin_brevan, welcome to dask community!
I don’t think this comes from dask, but more probably from the operating system dask is running in. Do you have access to the virtual machine or the docker image by any chance?
Hi, thank you for your response!
I made the docker image myself. It is based on a distroless debian image. When I have the image run:
print(dt.datetime.now())
it correctly prints the current time, in UTC:
2025-08-22 16:00:25.838473
Did you try this on Fargate? It’s really weird that Dask Scheduler doesn’t pick the OS date…
Yeah, on Fargate as well, the time is correct:
August 22, 2025 at 12:08 |
{"now": "2025-08-22T16:08:57.659969+00:00", "event": "dask.scheduler.start", "level": "info", "timestamp": "2025-08-22T16:08:57.660012Z UTC", "file": "dask.py", "line": 30} |
|---|
This log is printed from the same script that starts the distributed.Scheduler.
Which Dask version are you using? Does it occurs also on another environment like on your laptop?
The version on Fargate and in my local images:
dask[delayed,diagnostics,distributed,test]==2025.7.0
I also tried without Docker with the following version, in WSL Ubuntu:
dask, version 2023.12.1+dfsg
and I am getting the same issue, on my desktop, no images.
For information, issue has been identified, I encourage yout to follow up on github: Task Stream dashboard now showing date · Issue #9055 · dask/distributed · GitHub
Amazing, thank you. My org unfortunately has a policy that doesn’t let me interact with open source github repos, but I can view the issue, just not comment on it.