Lindsay Mathieson
2016-Apr-14 02:53 UTC
[Gluster-users] 3.7.9 Shards - strict-write-ordering
With 3.7.6 (and 3.7.7?) with sharded volumes we had to set: performance.strict-write-ordering: on otherwise we got wildly inconsistent file sizes between node mounts. However I've been rechecking it with 3.7.9 and that no longer seems to be the case, whcih is good as strict-write-ordering was quite a performance hit. Is this the case? has the issue been fixed for 3.7.9? thanks. ps. - also we no longer seem to need performance.stat-prefetch: off when live migrating VM's (used top cause data corruption). -- Lindsay
Lindsay Mathieson
2016-Apr-14 02:57 UTC
[Gluster-users] 3.7.9 Shards - strict-write-ordering
On 14 April 2016 at 12:53, Lindsay Mathieson <lindsay.mathieson at gmail.com> wrote:> With 3.7.6 (and 3.7.7?) with sharded volumes we had to set: > performance.strict-write-ordering: on > otherwise we got wildly inconsistent file sizes between node mounts. > However I've been rechecking it with 3.7.9 and that no longer seems to > be the case, whcih is good as strict-write-ordering was quite a > performance hit. > > Is this the case? has the issue been fixed for 3.7.9?ps. If strict-write-ordering is no longer needed can I just turn it off? or would I have to recreate the files on the volume? thanks, -- Lindsay