Many HDF5-DIAG warnings when used with xarray and dask

[nbud]

Steps to reproduce:

• Environment: Python 3.10, Ubuntu 22.04, the packages were installed in a new clean virtual environment using pip install netcdf4 dask xarray
• Create a new netcdf4 dataset, read it with xarray and dask; see example script below.
• Expected behavior: the script runs without warning
• Actual behavior: the script runs successfully but with many HDF5-DIAG warnings

pip list:

Package            Version
------------------ --------
cftime             1.6.2
click              8.1.3
cloudpickle        2.2.1
dask               2023.3.2
fsspec             2023.3.0
importlib-metadata 6.1.0
locket             1.0.0
netCDF4            1.6.3
numpy              1.24.2
packaging          23.0
pandas             1.5.3
partd              1.3.0
pip                23.0.1
python-dateutil    2.8.2
pytz               2023.3
PyYAML             6.0
setuptools         67.4.0
six                1.16.0
toolz              0.12.0
wheel              0.38.4
xarray             2023.3.0
zipp               3.15.0

Example script (taken from https://github.com/pydata/xarray/issues/7549#issue-1596115847):

import argparse
import pathlib
import tempfile
from typing import List

import netCDF4
import xarray

HERE = pathlib.Path(__file__).parent


def add_arguments(parser: argparse.ArgumentParser):
    parser.add_argument('count', type=int, default=200, nargs='?')
    parser.add_argument('--file-cache-maxsize', type=int, required=False)


def main():
    parser = argparse.ArgumentParser()
    add_arguments(parser)
    opts = parser.parse_args()

    if opts.file_cache_maxsize is not None:
        xarray.set_options(file_cache_maxsize=opts.file_cache_maxsize)

    temp_dir = tempfile.mkdtemp(dir=HERE, prefix='work-dir-')
    work_dir = pathlib.Path(temp_dir)
    print("Working in", work_dir.name)

    print("Making", opts.count, "datasets")
    dataset_paths = make_many_datasets(work_dir, count=opts.count)

    print("Combining", len(dataset_paths), "datasets")
    dataset = xarray.open_mfdataset(dataset_paths, lock=False)
    dataset.to_netcdf(work_dir / 'combined.nc')


def make_many_datasets(
    work_dir: pathlib.Path,
    count: int = 200
) -> List[pathlib.Path]:
    dataset_paths = []
    for i in range(count):
        variable = f'var_{i}'
        path = work_dir / f'{variable}.nc'
        dataset_paths.append(path)
        make_dataset(path, variable)

    return dataset_paths


def make_dataset(
    path: pathlib.Path,
    variable: str,
) -> None:
    ds = netCDF4.Dataset(path, "w", format="NETCDF4")
    ds.createDimension("x", 1)
    var = ds.createVariable(variable, "i8", ("x",))
    var[:] = 1
    ds.close()


if __name__ == '__main__':
    main()

Content of stdout:

Working in work-dir-kt4qcsng
Making 200 datasets
Combining 200 datasets

Content of stderr (partial):

HDF5-DIAG: Error detected in HDF5 (1.12.2) thread 1:
  #000: H5A.c line 528 in H5Aopen_by_name(): can't open attribute
    major: Attribute
    minor: Can't open object
  #001: H5VLcallback.c line 1091 in H5VL_attr_open(): attribute open failed
    major: Virtual Object Layer
    minor: Can't open object
  #002: H5VLcallback.c line 1058 in H5VL__attr_open(): attribute open failed
    major: Virtual Object Layer
    minor: Can't open object
  #003: H5VLnative_attr.c line 130 in H5VL__native_attr_open(): can't open attribute
    major: Attribute
    minor: Can't open object
  #004: H5Aint.c line 545 in H5A__open_by_name(): unable to load attribute info from object header
    major: Attribute
    minor: Unable to initialize object
  #005: H5Oattribute.c line 494 in H5O__attr_open_by_name(): can't locate attribute: '_QuantizeBitGroomNumberOfSignificantDigits'
    major: Attribute
    minor: Object not found
HDF5-DIAG: Error detected in HDF5 (1.12.2) thread 1:
  #000: H5A.c line 528 in H5Aopen_by_name(): can't open attribute
    major: Attribute
    minor: Can't open object
  #001: H5VLcallback.c line 1091 in H5VL_attr_open(): attribute open failed
    major: Virtual Object Layer
    minor: Can't open object
  #002: H5VLcallback.c line 1058 in H5VL__attr_open(): attribute open failed
    major: Virtual Object Layer
    minor: Can't open object
  #003: H5VLnative_attr.c line 130 in H5VL__native_attr_open(): can't open attribute
    major: Attribute
    minor: Can't open object
  #004: H5Aint.c line 545 in H5A__open_by_name(): unable to load attribute info from object header
    major: Attribute
    minor: Unable to initialize object
  #005: H5Oattribute.c line 494 in H5O__attr_open_by_name(): can't locate attribute: '_QuantizeGranularBitRoundNumberOfSignificantDigits'
    major: Attribute
    minor: Object not found
HDF5-DIAG: Error detected in HDF5 (1.12.2) thread 1:
  #000: H5A.c line 528 in H5Aopen_by_name(): can't open attribute
    major: Attribute
    minor: Can't open object
  #001: H5VLcallback.c line 1091 in H5VL_attr_open(): attribute open failed
    major: Virtual Object Layer
    minor: Can't open object
  #002: H5VLcallback.c line 1058 in H5VL__attr_open(): attribute open failed
    major: Virtual Object Layer
    minor: Can't open object
  #003: H5VLnative_attr.c line 130 in H5VL__native_attr_open(): can't open attribute
    major: Attribute
    minor: Can't open object
  #004: H5Aint.c line 545 in H5A__open_by_name(): unable to load attribute info from object header
    major: Attribute
    minor: Unable to initialize object
  #005: H5Oattribute.c line 494 in H5O__attr_open_by_name(): can't locate attribute: '_QuantizeBitRoundNumberOfSignificantBits'
    major: Attribute
    minor: Object not found
HDF5-DIAG: Error detected in HDF5 (1.12.2) thread 2:
  #000: H5A.c line 528 in H5Aopen_by_name(): can't open attribute
    major: Attribute
    minor: Can't open object
...

Downgrading to netcdf4=1.5.8 fixes the issue. I don't know if the root cause is netcdf4, xarray or dask, I just wanted to raise awareness here. Probably related: https://github.com/Unidata/netcdf4-python/issues/1241

[ocefpaf]

I was able to reproduce this with conda packages. Only when downgrading hdf5 to 1.12.1 the outputs went away :-/

[bluppfisk]

Is there a solution to this yet? The massive diag output (40k lines at a time) makes it really difficult to debug applications built with netcdf4