dask / dask/distributed

Find the future/partition/file that caused error in Pandas operations

Open
#6,209 2 comments 0 reactions 0 assignees View on GitHub
Dominant language
Python
Stars
1.7k
Forks
778
Avg merge
2h 50m
Merged PRs (30d)
3

Description

I'm getting a `KilledWorker` error when doing `ddf.set_index()` on a large dataset (700 million rows, 4k partitions/files). It's reproducible just by doing

```python
data_ddf = dd.read_csv("common_file_pattern*.tsv.gz")
data_ddf = data_ddf.set_index("index_col")`
```

Error:
```python
KilledWorker: ("('assign-7d4fa5a59adb2ffc98d6a3d88ef328bf', 328)",
)
```

I suspect it's an error on the task itself and not the cluster (*). Per [diagnostics guidelines here](https://distributed.dask.org/en/latest/killed.html?highlight=killed%20worker#unrecoverable-exception) I try to call `client.recreate_error_locally()` but need the future that failed. I cannot find the future `('assign-7d4fa5a59adb2ffc98d6a3d88ef328bf', 328)` in `client.futures`. Where can I find the future or partition/file (since each partition is a file here) that it failed on? If this is not possible then I'll have to go through the failed worker's logs, which can be alot.

(*) The error is consistent across different cluster configurations (I have 1k nodes with 45 CPUs each and tried giving each worker more processes etc). The task ran for a while (50% completed) and errored out every time so I suspect a partition/file is causing the problem.

Contributor guide

Open the contributing guide

Assessment

This issue has not been assessed yet.

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.