Re: Replication failure, slave requesting old segments

Adrian Klaver wrote:
> On 08/12/2018 03:54 PM, Stephen Frost wrote:
>> Greetings,
>> 
>> * Phil Endecott (spam_from_pgsql_lists@chezphil.org) wrote:
>>> OK.  I think this is perhaps a documentation bug, maybe a missing
>>> warning when the master reads its configuration, and maybe (as you say)
>>> a bad default value.
>> 
>> If we consider it to be an issue worthy of a change then we should
>> probably just change the default value, and maybe not even allow it to
>> be set lower than '1'.
>> 
>
> I would say leave the default at 0 as it leaves no doubt that you are 
> performing without a net. A setting of '1' implies you are covered and 
> for a fast moving cluster or slow moving one with sufficient downtime 
> that would not be the case.

Can you explain how it can fail in the case of a "slow moving cluster with
sufficient downtime"?

It seems to me that if I have correctly understood what happened in this
case then 0, the default, really cannot ever work properly when you have
enabled WAL archiving plus streaming.

> Better to let the end user know this is not 
> a simple problem and some thought needs to go into configuration.

I certainly agree that this is "not a simple problem", having read something
like 18,000 words of documentation multiple times and, apparently, still
got it wrong in multiple ways.


Regards, Phil.