On 2015-05-08 22:08:31 -0400, Robert Haas wrote:
> That seems a bit better. I think it's really important, if we're
> going to start to try to make fsync=off anything other than a toy,
I think it's long past that. I've seen many, many people use it during
initial data loading.
> that we document really clearly the circumstances in which it is or is
> not safe:
Yea, we really should have done that a long time ago.
> - If you crash while fsync=off, your cluster may be corrupted.
HW crash, right?
> - If you crash while fsync=on, but it was off at the last checkpoint,
> your cluster may be corrupted.
> - If you turn fsync=off, do stuff, turn fsync=on, and checkpoint
> successfully, a subsequent crash should not corrupt anything.
Yep.
> Of course, even the last one isn't totally bullet-proof. Suppose one
> backend fails to absorb the new setting for some reason...
I've a hard time worrying much about that one...
Greetings,
Andres Freund