Let’s say there are two python scripts that want to write data to the same table which is stored in an SQLite file using the sqlite3 module. The SQLite-file is stored on an NFS filesystem. In the SQLite-FAQ I read:
SQLite uses reader/writer locks to control access to the database.
[…] But use caution: this locking mechanism might not work correctly
if the database file is kept on an NFS filesystem. This is because
fcntl() file locking is broken on many NFS implementations. You should
avoid putting SQLite database files on NFS if multiple processes might
try to access the file at the same time.
Does this mean it is not possible at all or is there some way to ensure that one process waits until the other is done?
The INSERTs are not complex. Just some:
INSERT_STATEMENT = "INSERT INTO some_table (row, col, val) VALUES (?, ?, ?)"
connection.executemany(INSERT_STATEMENT, triples)
And the inserted sets are disjoint.
A further question: Does the NFS-Problems occure when two processes try to write to the same table or when they try to write to the same database (which is a file)? Would it be a workaround to let each process create its own table in the same database (file) and write to that?
Do not use SQLite with NFS. It is as simple as that. NFS semantics are different than regular filesystems and are looser. You will eventually get corruption. Every now and then someone on the SQLite-users mailing list posts with their “workarounds”. They never work although they appear to in the short term.