On Sat, 2013-04-06 at 16:44 +0200, Andres Freund wrote:
> I think we can just make up the rule that changing full page writes also
> requires SpinLockAcquire(&xlogctl->info_lck);. Then its easy enough. And
> it can hardly be a performance bottleneck given how infrequently its
> modified.
That seems like a good idea to me. As it stands, checksums basically
force full page writes to be on; so we should either fix that or
document it.
> In retrospect I think making up the rule that full_page_writes changes
> imply a checkpoint would have made things easier performance and
> codewise.
I don't even see why we allow changing that while the server is on.
Either the I/O system requires it for safety or not, right?
Regards,Jeff Davis
