Bug: Intermittent database corruption fails at read_snapshot_or_default

[D1plo1d]

1. expected result: Expected config.open() to open the database
2. actual result:

• config.open() some times fails with the following stack trace (note this is in reverse order due to journalctl): https://gist.github.com/D1plo1d/e1ac7171375f8fbd47c1aed9947b59d6
• once config.open() fails once it will continue to fail if retried indicating that the database is corrupted(?)
• I delete the database and it works for a while but then the problem re-occurs within less then a day.

3. sled version: 0.34.2 + some small error handling experiments: https://github.com/D1plo1d/sled/tree/53d0e30a749f27fb808101730f1794a5f85b6216
4. rustc version: rustc 1.44.1 (c7087fe00 2020-06-17)
5. operating system: Rasbian 10 (Buster)
6. minimal code sample that helps to reproduce the issue:

    let db_file = env::var("SLED_DB_PATH")
        .expect("$SLED_DB_PATH not set");

    let config = sled::Config::default()
        .path(db_file.clone())
        .cache_capacity(10_000_000);

    let db = config.open()
        .map(|db| Arc::new(db))
        .with_context(|| format!("Unable to open sled database: {}", db_file))?;

1. logs, panic messages, stack traces: https://gist.github.com/D1plo1d/e1ac7171375f8fbd47c1aed9947b59d6

[spacejam]

Thanks for reporting this! I believe I may have fixed it here, and I'm about to cut 0.34.4 which will include the fix.

[spacejam]

0.34.4 has been cut which I believe addresses this. please re-open if you encounter similar issues!