a qustion about the order of write wal and write manifest

[baobaomaomeng]

Consider the following code（in min-lsm/src/lsm_storage line714-line725）:

if self.options.enable_wal {
    std::fs::remove_file(self.path_of_wal(sst_id))?;
}
self.manifest
    .as_ref()
    .unwrap()
    .add_record(&state_lock, ManifestRecord::Flush(sst_id))?;

self.sync_dir()?;

If the program loses power after the WAL deletion is completed, an error will occur during recovery, and we will be unable to restore the flushed memtable（which is not flush in record）. Therefore, in theory, we should reverse the order of these two.

However, I still have a concern regarding the reordering of code and instructions by the compiler and CPU. The write operation to the WAL (Write - Ahead Log) and the write operation to the manifest do not establish a happen - before relationship. In theory, their execution order can be swapped. If this happens, we may encounter the problem described in the above - mentioned code once again.

[skyzh]

Yep, I agree -- I'll fix this in the next edition of the course.