Re: Return value of pg_promote()

On Thu, Jun 08, 2023 at 04:53:50PM +0530, Ashutosh Sharma wrote:
> Thanks for sharing your thoughts, Laurenz and Fujii-san. I've prepared
> a patch that makes pg_promote error out if it couldn't send SIGUSR1 to
> the postmaster or if the postmaster died in the middle of standby
> promotion. PFA. Please note that now (with this patch) pg_promote only
> returns false if the standby could not be promoted within the given
> wait time. In case of any kind of failure, it just reports an error
> based on the type of failure that occurred.

if (kill(PostmasterPid, SIGUSR1) != 0)
{
- ereport(WARNING,
- (errmsg("failed to send signal to postmaster: %m")));
(void) unlink(PROMOTE_SIGNAL_FILE);
- PG_RETURN_BOOL(false);
+ ereport(ERROR,
+ (errmsg("failed to send signal to postmaster: %m")));
}

Shouldn't you assign an error code to this one rather than the
default one for internal errors, like ERRCODE_SYSTEM_ERROR?

/* return immediately if waiting was not requested */
@@ -744,7 +743,9 @@ pg_promote(PG_FUNCTION_ARGS)
* necessity for manual cleanup of all postmaster children.
*/
if (rc & WL_POSTMASTER_DEATH)
- PG_RETURN_BOOL(false);
+ ereport(FATAL,
+ (errcode(ERRCODE_ADMIN_SHUTDOWN),
+ errmsg("terminating connection due to unexpected postmaster exit")));

I would add an errcontext here, to let somebody know that the
connection died while waiting for the promotion to be processed, say
"while waiting on promotion".
--
Michael