Unsuccessful recovery from ConnectionError

[jhamman]

I'm trying to push a large-ish dataset to GCS via xarray/dask/zarr/gcsfs. Things are generally working during the setup and for the first part of the upload. However, after a bit, I'm getting a ConnectionError that is not recoverable. I'm pushing from a server at the University of Washington to bucket at "US-CENTRAL1". I would image the network at UW is pretty stable.

Version info:

xarray: jhamman:fix/zarr_set_attrs dask: 0.16.0 zarr: master gcsfs: master

Details of the full traceback are in this gist: https://gist.github.com/jhamman/25ddda993ad5b768e4b8289904be6779

cc @mrocklin, @martindurant

xref: https://github.com/pangeo-data/pangeo/issues/48, https://github.com/pydata/xarray/pull/1800

[mrocklin]

Out of curiosity was using retries= in dask effective?

[jhamman]

@mrocklin - yes, retries=10 did seem to solve this. Perhaps that's all we need here. https://github.com/pydata/xarray/pull/1811 may be able to close this issue for me.

[mrocklin]

We should still resolve connection issues within gcsfs.

On Mon, Jan 8, 2018 at 11:57 AM, Joe Hamman notifications@github.com wrote:

@mrocklin https://github.com/mrocklin - yes, retries=10 did seem to solve this. Perhaps that's all we need here. pydata/xarray#1811 https://github.com/pydata/xarray/pull/1811 may be able to close this issue for me.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/dask/gcsfs/issues/61#issuecomment-356043149, or mute the thread https://github.com/notifications/unsubscribe-auth/AASszO7aySCg_10hzxVExVtcwNTcXY8eks5tIlb0gaJpZM4RVaTU .

[yan-hic]

POST requests were not wrapped in a retry logic back then but they are now (just). @jhamman can this be closed ?