BUG: Failed wheel upload for h5py

[larsoner]

After most of the h5py binaries uploading successfully last night the 3.11 one failed:

https://github.com/h5py/h5py/actions/runs/8445531348/job/23133109268#step:5:408

Warning:  Distribution "h5py-3.10.0-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl" already exists. Removing.
Creating package "h5py"
Creating release "3.10.0"
Uploading file "scientific-python-nightly-wheels/h5py/3.10.0/h5py-3.10.0-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl"

  0%|          | 0.00/5.17M [00:00<?, ?B/s]
  0%|          | 16.0k/5.17M [00:00<01:04, 83.8kB/s]
  2%|▏         | 128k/5.17M [00:00<00:10, 507kB/s]  
  9%|▉         | 496k/5.17M [00:00<00:03, 1.63MB/s]
 22%|██▏       | 1.14M/5.17M [00:00<00:01, 3.26MB/s]
 61%|██████▏   | 3.17M/5.17M [00:00<00:00, 8.93MB/s]
5.18MB [00:00, 6.13MB/s]                            
standard python located at:
  https://anaconda.org/scientific-python-nightly-wheels/h5py

Error:  ("Invalid dist_id ObjectId('6603800020b28178c3fabc66') does not exist", [404](https://github.com/h5py/h5py/actions/runs/8445531348/job/23133109268#step:5:405))

Any ideas on what the problem could? Could the upload be made more robust somehow? It removed the wheel built the previous night prior to trying to upload the new wheel, so I ended up with a failed --pre install on a Linux 3.11 CI.

[matthewfeickert]

@larsoner Looks like you didn't try to rerun the workflow. Please do that (by which I mean just select the rerun button from the GitHub UI).

Could the upload be made more robust somehow?

Not really as it is just calling the anaconda-client CLI API

https://github.com/scientific-python/upload-nightly-action/blob/535c72c70a11e52673bee0471be1ca52161a4b5a/cmd.sh#L75-L79

I'd suggest opening this same issue but with anaconda-client: https://github.com/Anaconda-Platform/anaconda-client/issues

[larsoner]

No permissions to restart myself so I'll probably just wait 24h for the next nightly :)

Closing since it does seem to be an upstream bug!

[matthewfeickert]

https://github.com/Anaconda-Platform/anaconda-client/issues/712

Ah okay. Can we ping @tacaswell or someone else with permissions to restart it though? It would be nice to not have missing wheels that could potentially hit others with your same bug.

Closing since it does seem to be an upstream bug!

Thanks! I've opened up https://github.com/Anaconda-Platform/anaconda-client/issues/712 for this.

[larsoner]

Yeah I think @tacaswell or @takluyver should be able to restart this job (copied from above for convenience) to fix the incomplete nightly wheel upload:

https://github.com/h5py/h5py/actions/runs/8445531348/job/23133109268#step:5:408

[takluyver]

I've hit restart, it seems to have succeeded now. :+1: