Re: [BUG] pg_basebackup from disconnected standby fails

On Wed, Oct 26, 2016 at 12:39 PM, Michael Paquier
<michael(dot)paquier(at)gmail(dot)com> wrote:
> On Wed, Oct 26, 2016 at 1:40 PM, Amit Kapila <amit(dot)kapila16(at)gmail(dot)com> wrote:
>> If you are inclined towards this solution, then I think what we need
>> to do is to change the API UpdateMinRecoveryPoint() such that it's
>> second parameter can take three values. 0 means update
>> minRecoveryPoint to passed lsn if minRecoveryPoint < lsn; 1 means
>> update minRecoveryPoint to latest replayed point if minRecoveryPoint <
>> lsn, same as currently false for *force*; 2 indicates same behaviour
>> as current *force* as true. Also we need to pass currentTLI parameter
>> (lastCheckPoint.ThisTimeLineID) to this API to update
>> minRecoveryPointTLI. I have not tried this, but I think something on
>> these lines should work.
>
> 0 is the behavior that you get when force = false, so it works as 1, no?
>

I don't think so. Refer below code:

UpdateMinRecoveryPoint()
{
..
else if (force || minRecoveryPoint < lsn)
{
/*
* To avoid having to update the control file too often, we update it
* all the way to the last record being replayed, even though 'lsn'
* would suffice for correctness. This also allows the 'force' case
* to not need a valid 'lsn' value.
..
}

If my understanding is right, then the changes proposed by you as
below is not what is intended here. I think you need to do something
as I have mentioned above.

+ /*
+ * Update minRecoveryPoint just past the last redo checkpoint if
+ * necessary. This ensures that at next startup minRecoveryPoint will
+ * not be past the next point it would start at, preventing any
+ * weird behaviors with for example backups taken from standbys that
+ * rely on minRecoveryPoint as end backup location.
+ */
+ UpdateMinRecoveryPoint(RedoRecPtr, false);
+

--
With Regards,
Amit Kapila.
EnterpriseDB: http://www.enterprisedb.com