openml.org is down

[Alkem0s]

I haven't been able to connect to the openml website for the past two days. I've had no problems before now.
I currently get an ERR_CONNECTION_TIMED_OUT.
The Python API also does not work because of this when trying to download datasets.
I tried accessing it from different devices and using a proxy as well as a vpn but it did not solve the problem.
I do not think it is on my end but some help would be appreciated.

[p-wojciechowski]

I have the same problem too. It seems it is down for everyone: https://www.isitdownrightnow.com/openml.org.html

[Hanxiao-coder]

Good to know it's not just me...

[nikiburggraf]

Joined the openml Slack because I also ran into this and found out the following -- it appears that the root cause is a cyberattack on TU/e, the university that hosts the website, which led to the university shutting its network down. Seems like the network is starting to come back online, but they are not sure when openML itself will be

https://www.tue.nl/en/news-and-events/cyberattack

[joaquinvanschoren]

Indeed, OpenML is currently unreachable because of a cyberattack that hit TU Eindhoven. OpenML itself is not affected and we're in contact with the university IT team to bring it back soon.

In the meantime, we are preparing to set up a secondary server (on an entirely different network) in the Dutch supercomputing center so that such an outage won't happen again.

[cacti77]

This is weird, because I can run these lines right now successfully in Python 3.9 on Windows:

        from sklearn.datasets import fetch_openml
        heart = fetch_openml(data_id=336, parser='auto', n_retries=10, delay=5.0)
        print(heart.details)

but the fetch_openml() line fails with this from a Microsoft-hosted Windows agent in Azure DevOps:

urllib.error.URLError: <urlopen error [WinError 10060] A connection attempt failed because the connected party did not properly respond after a period of time, or established connection failed because connected host has failed to respond>

I don't understand how it can run successfully locally if the whole OpenML network is down.

[joaquinvanschoren]

OpenML automatically caches datasets so if you have downloaded it earlier it will still work.

[cacti77]

OpenML automatically caches datasets so if you have downloaded it earlier it will still work.

Thanks, yes, if I try downloading a different dataset, or delete the scikit_learn_data folder under my Windows user account then I can't download anything.

[ExtraFlash]

Do you know when the website will be available?

[PGijsbers]

The university is taking longer than expected to bring the servers back up, which is why we are now trying to host a read-only server outside of the TUe, so that users can at least access datasets and tasks as usual. We do not know how long it will take (this too requires some cooperation from already very busy IT staff at TUE), but we are doing our best to get it up as soon as we can.

Regular servers are unlikely to be operational before next week, we do not have a very clear timeline. It is unfortunately out of our control.

[BradKML]

Welp that is a little sad, but has anyone considered making mirrors of the same service on other websites/servers as well so that it can stay up? Not saying IPFS is the best thing since sliced bread but it could work? @PGijsbers @joaquinvanschoren

[joaquinvanschoren]

We're setting up a mirror right now (we just got internal access over the weekend), hopefully more news later today.

[joaquinvanschoren]

Please see this discussion board for context and updates on the OpenML outage: https://github.com/orgs/openml/discussions/20

[joaquinvanschoren]

Since OpenML is fully operational again I'll close this issue.
Thanks everyone for reporting and helping us, and apologies again for the inconvenience.