[nycbug-talk] Approach for the NFS cluster

Miles Nordin carton at Ivy.NET
Wed May 20 10:05:43 EDT 2009


>>>>> "il" == Isaac Levy <isaac at diversaform.com> writes:

    il> I can see potential problems in data loss- (loosing the box as
    il> file(s) are being written, or were recently written).

Even with ZIL disabled any data written more than 30 seconds ago is safe.

If ZFS recovers at all, ZFS will always recover to some state it
passed through prior to the crash (modulo fsync() which I think does
fuck with the timeline).

Disabling the ZIL does not at all increase the likelihood that ZFS
won't recover at all (pool loss).

If you are not using fsync() or NFS, the ZIL is never even written at
all.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 304 bytes
Desc: not available
URL: <http://lists.nycbug.org/pipermail/talk/attachments/20090520/3dfc44f1/attachment.bin>


More information about the talk mailing list