Replies: 1 comment 3 replies
-
|
this is a long-standing issue by now (see e.g. #7079). Note that this only happens with a threading scheduler, so you can use |
Beta Was this translation helpful? Give feedback.
-
|
Thanks for pointing to that issue, it is reassuring to see this is not a quirk of my setup. Lock Error Stack Trace File "HIDDEN/lib/python3.12/site-packages/xarray/core/dataset.py", line 2327, in to_netcdf
return to_netcdf( # type: ignore # mypy cannot resolve the overloads:(
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "HIDDEN/lib/python3.12/site-packages/xarray/backends/api.py", line 1346, in to_netcdf
writes = writer.sync(compute=compute)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "HIDDEN/lib/python3.12/site-packages/xarray/backends/common.py", line 267, in sync
delayed_store = chunkmanager.store(
^^^^^^^^^^^^^^^^^^^
File "HIDDEN/lib/python3.12/site-packages/xarray/namedarray/daskmanager.py", line 249, in store
return store(
^^^^^^
File "HIDDEN/lib/python3.12/site-packages/dask/array/core.py", line 1229, in store
compute_as_if_collection(Array, store_dsk, map_keys, **kwargs)
File "HIDDEN/lib/python3.12/site-packages/dask/base.py", line 403, in compute_as_if_collection
return schedule(dsk2, keys, **kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "HIDDEN/lib/python3.12/site-packages/dask/multiprocessing.py", line 234, in get
result = get_async(
^^^^^^^^^^
File "HIDDEN/lib/python3.12/site-packages/dask/local.py", line 500, in get_async
fire_tasks(chunksize)
File "HIDDEN/lib/python3.12/site-packages/dask/local.py", line 482, in fire_tasks
dumps((dsk[key], data)),
^^^^^^^^^^^^^^^^^^^^^^^
File "HIDDEN/lib/python3.12/site-packages/cloudpickle/cloudpickle.py", line 1479, in dumps
cp.dump(obj)
File "HIDDEN/lib/python3.12/site-packages/cloudpickle/cloudpickle.py", line 1245, in dump
return super().dump(obj)
^^^^^^^^^^^^^^^^^
File "HIDDEN/lib/python3.12/multiprocessing/synchronize.py", line 101, in __getstate__
context.assert_spawning(self)
File "HIDDEN/lib/python3.12/multiprocessing/context.py", line 374, in assert_spawning
raise RuntimeError(
RuntimeError: Lock objects should only be shared between processes through inheritance |
Beta Was this translation helpful? Give feedback.
-
|
for cluster = LocalCluster(..., processes=True, ...)
client = Client(cluster)(I don't think we've been testing the local scheduler, so it doesn't surprise me too much that using that doesn't work too well. I might be wrong, though) |
Beta Was this translation helpful? Give feedback.
-
|
It does indeed look as if there is a bug when using |
Beta Was this translation helpful? Give feedback.
-
Hi all,
I am using
xarrayanddaskat scale, and have been running into issues with occasional complete segmentation faults. As in, some times on the same data and code it will fail, and other times it succeeds.Occasionally, it will report more helpful information, like
NetCDF- HDF error,malloc,free pointer, but other times no such luck.The same code succeeds every time when running without a local dask cluster, but when a dask cluster is present, failures occur.
I am looking for any help / tips on diagnosing and identifying the cause of these errors. It clearly has something to do with dask, its worker setup and netcdf, but further then that I am not sure.
Example Error
These errors are occurring when loading large > 50 GB datasets from disk, modifying the shape, and values, and saving again to disk.
All files are netcdf compliant, and are being loaded with
xr.open_mfdatasetand saved with.to_netcdf.Result of `xr.show_versions()`
What I have tried
.compute(scheduler='single-threaded').compute(rerun_exceptions_locally=True)Beta Was this translation helpful? Give feedback.
All reactions