couchdb-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Apache Wiki <wikidi...@apache.org>
Subject [Couchdb Wiki] Update of "Durability Matrix" by AdamKocoloski
Date Thu, 17 Sep 2009 23:59:49 GMT
Dear Wiki user,

You have subscribed to a wiki page or wiki category on "Couchdb Wiki" for change notification.

The following page has been changed by AdamKocoloski:
http://wiki.apache.org/couchdb/Durability_Matrix

------------------------------------------------------------------------------
  
  = How file:sync() works =
  
- All data flushing is performed by file:sync(IoDevice). This function maps to fsync(fd) on
most platforms, with the notable exception of OS/X where fcntl(F_FULLSYNC) is called instead.
+ All data flushing is performed by file:sync(IoDevice). This function maps to fsync(fd) on
most platforms, with the notable exception of OS/X where fcntl(F_FULLFSYNC) is called instead.
  
- A call to fsync(fd) should ensure, on completion, that all pending writes have reached the
disk. On many platforms, and Linux in particular, this promise is not upheld in the presence
of write-caching disk controllers. On affected configurations, a call to fsync(fd) will complete
as soon as all pending writes have reached the cache of the disk controller itself. If your
controller has battery-backing, you might decide this is sufficiently durable. On OS/X, the
fcntl(F_FULLSYNC) really gets all pending writes to disk even in the presence of a write-caching
disk controller. This is the reason that CouchDB is "slower" on OS/X.
+ A call to fsync(fd) should ensure, on completion, that all pending writes have reached the
disk. On many platforms, and Linux in particular, this promise is not upheld in the presence
of write-caching disk controllers. On affected configurations, a call to fsync(fd) will complete
as soon as all pending writes have reached the cache of the disk controller itself. If your
controller has battery-backing, you might decide this is sufficiently durable. On OS/X, the
fcntl(F_FULLFSYNC) really gets all pending writes to disk even in the presence of a write-caching
disk controller. This is the reason that CouchDB is "slower" on OS/X.
  
  = How CouchDB uses fsync() =
  

Mime
View raw message