The following is cited from the zookeeper book by Flavio Junqueira and Benjamin Reed:
dataDir is the directory where the fuzzy snapshots of the in-memory database willbe stored. If this server is part of an ensemble, the myid file will also be in this directory. The dataDir does not need to reside on a dedicated device. The snapshots are written using a background thread that does not lock the database, and the writes to storage are not synced until the snapshot is complete. I don't quiet understand The snapshots are written using a background thread that does not lock the database.Does it mean that the in memory database is not locked while the background thread reads it and writes the snapshot to dataDir? If this is the case, then if the in-memory data keeps changing while the background thread reads it, then there must be some mechanism for the snapshot to write completely and correctly(one scenario is the data that the thread has just read,but the in-memory database has just changed it). Also I don't understand the writes to storage are not synced until the snapshot is complete, what's the process of writing to storage, snapshot complets,sync. Can someone explain more?Thanks.
