|
Description
|
See comments.
xxxxx@xxxxx.com 2005-06-05 19:54:42 GMT
Jeff a while back proposed we add a mount option to override the
synchronous data (& metadata) demands on an file system.
The options and semantics proposed for a new mount option were:
sync=default
As today, we would flush file system transactions out to
stable storage (the intent log) on demand (fsync, sync(1M)
and O_DSYNC, O_RSYNC, etc)
sync=always
For the ultra cautious, every file system transaction would
be flushed to stable storage on system call return.
This obviously has a big performance penalty.
sync=delayed
All file system transactions only commit to stable storage
on the next DMU transaction group commit. This is currently
every 5 seconds. This option will give the highest performance.
However, it is very dangerous as ZFS would be ignoring
the synchronous transaction demands of applications and NFS.
|