Berkeley DB Reference Guide: Environment

Remote filesystems

When Berkeley DB database environment shared memory regions are backed by the filesystem, it is a common application error to create database environments backed by remote filesystems such as the Network File System (NFS), Windows network shares (SMB/CIFS) or the Andrew File System (AFS). Remote filesystems rarely support mapping files into process memory, and even more rarely support correct semantics for mutexes if the mapping succeeds. For this reason, we recommend database environment directories be created in a local filesystem.

For remote filesystems that do allow remote files to be mapped into process memory, database environment directories accessed via remote filesystems cannot be used simultaneously from multiple clients (that is, from multiple computers). No commercial remote filesystem of which we're aware supports coherent, distributed shared memory for remote-mounted files. As a result, different machines will see different versions of these shared region files, and the behavior is undefined.

Databases, log files, and temporary files may be placed on remote filesystems, as long as the remote filesystem fully supports standard POSIX filesystem semantics (although the application may incur a performance penalty for doing so). Further, read-only databases on remote filesystems can be accessed from multiple systems simultaneously. However, it is difficult (or impossible) for modifiable databases on remote filesystems to be accessed from multiple systems simultaneously. The reason is the Berkeley DB library caches modified database pages, and when those modified pages are written to the backing file is not entirely under application control. If two systems were to write database pages to the remote filesystem at the same time, database corruption could result. If a system were to write a database page back to the remote filesystem at the same time as another system read a page, a core dump in the reader could result.

FreeBSD note:

Some historic FreeBSD releases will return ENOLCK from fsync and close calls on NFS-mounted filesystems, even though the call has succeeded. To support Berkeley DB on these releases, the Berkeley DB code should be modified to ignore ENOLCK errors, or no Berkeley DB files should be placed on NFS-mounted filesystems on these systems. Note that current FreeBSD releases do not suffer from this problem.

Linux note:

Some historic Linux releases do not support complete semantics for the POSIX fsync call on NFS-mounted filesystems. No Berkeley DB files should be placed on NFS-mounted filesystems on these systems. Note that current Linux releases do not suffer from this problem.

Copyright (c) 1996-2005 Sleepycat Software, Inc. - All rights reserved.