sqlite3.OperationalError database or disk is full on a new directory #349
Description
I am running diskcache in docker to maintain cache between multiple dask workers.
In the beginning of the execution I remove the directory (cleaning) and then create new one (fresh start)
shutil.rmtree(self._cache_directory)
os.makedirs(self._cache_directory, exist_ok=True)
I have logs to verify that directory was indeed deleted and created anew.
Immediately after this run diskcache.Cache(directory)
Expected Behavior
The cache should initialize successfully on the freshly created directory.
Actual Behavior
Intermittently receive a SQLite "database or disk is full" error during initialization:
"exception": {
"args": [
"database or disk is full"
],
"class": "sqlite3.OperationalError",
"source": [
{
"file": "",
"function": "set_contract_path_cache",
"line": 148,
"text": "_CACHE = diskcache.Cache(directory)"
},
{
"file": "",
"function": "__init__",
"line": 509,
"text": "self.reset(key, value, update=False)"
},
{
"file": "",
"function": "reset",
"line": 2467,
"text": "sql('PRAGMA %s = %s' % (pragma, value)).fetchall()"
},
{
"file": "",
"function": "execute",
"line": 32,
"text": "return self.cursor().execute(*args, **kwargs)"
},
]```
I have checked, there is more than enough disk space free. I am confused why would I see this error on a empty directory.
### Environment
- diskcache version: 4.1.0
- Python version: 3.9
- Operating system: Linux (in Docker container)
- Docker: Yes, running multiple dask workers sharing the same cache directory
### Additional Context
- The error is intermittent and doesn't happen every time
- I've verified there is plenty of disk space available
- The directory is freshly created before initializing the cache
- The error happens during the PRAGMA statement execution in the `reset` method
Any guidance on how to resolve or work around this issue would be greatly appreciated.