On Wed, Dec 31, 2014 at 12:22 AM, Alexey Vasiliev <leopard_ne@inbox.ru> wrote:
> Tue, 30 Dec 2014 21:39:33 +0900 от Michael Paquier <michael.paquier@gmail.com>:
> As I understand now = (pg_time_t) time(NULL); return time in seconds, what is why I multiply value to 1000 to compare
withrestore_command_retry_interval in milliseconds.
This way of doing is incorrect, you would get a wait time of 1s even
for retries lower than 1s. In order to get the feature working
correctly and to get a comparison granularity sufficient, you need to
use TimetampTz for the tracking of the current and last failure time
and to use the APIs from TimestampTz for difference calculations.
> I am not sure about small retry interval of time, in my cases I need interval bigger 5 seconds (20-40 seconds). Right
nowI limiting this value be bigger 100 milliseconds.
Other people may disagree here, but I don't see any reason to put
restrictions to this interval of time.
Attached is a patch fixing the feature to use TimestampTz, updating as
well the docs and recovery.conf.sample which was incorrect, on top of
other things I noticed here and there.
Alexey, does this new patch look fine for you?
--
Michael