Re: O_DSYNC broken on MacOS X?

On Oct 7, 2010, at 12:26 PM, Robert Haas wrote:

> On Thu, Oct 7, 2010 at 11:45 AM, Greg Smith <greg(at)2ndquadrant(dot)com> wrote:
>> Robert Haas wrote:
>>> Proposed doc patch attached.
>>
>> Looks accurate to me. I like the additional linking to the Reliability page
>> you put in there too. Heavily referencing that important page from related
>> areas is a good thing, particularly now that it's got a lot more details
>> than it used to.
>
> Cool, thanks for the fast review. I suspect there are more details
> that could stand to be added to the WAL reliability page as well, but
> I don't know what they are so I can't add them.
>
> I still have the feeling that we have not put quite a large enough
> red, blinking light around this issue, but I don't have a concrete
> suggestion.

I think the general problem is that there is no simple way to verify that a PostgreSQL commit is pushing the bits to persistent storage. It would be helpful if there were a platform-specific, volume-specific tool to deduce this. Currently, there is no warning light that goes on when commits are not persistent.

On Linux, a tool could check filesystem parameters, hdparm (if relevant), and hard drive and controller specs (possibly against a blacklist of known liars).

Perhaps a simpler tool could run a basic fsyncs-per-second test and prompt the DBA to check that the numbers are within the realm of possibility.

How else can a DBA today ensure that a commit is a commit?

Cheers,
M