On 02.01.2011 15:41, Simon Riggs wrote:
> On Sat, 2011-01-01 at 23:36 -0500, Robert Haas wrote:
>> On Sat, Jan 1, 2011 at 6:54 AM, Simon Riggs<simon@2ndquadrant.com> wrote:
>>> Yes, working out the math is a good idea. Things are much clearer if we
>>> do that.
>>>
>>> Let's assume we have 98% availability on any single server.
>>>
>>> 1. Having one primary and 2 standbys, either of which can acknowledge,
>>> and we never lock up if both standbys fail, then we will have 99.9992%
>>> server availability. (So PostgreSQL hits "5 Nines", with data
>>> guarantees). ("Maximised availability")
>>
>> I don't agree with this math. ...(snip by Simon)... 99.96%.
>
> OK, so that is at least 99.96%. Cool.
>
> The key point here is not (1), but option (4).
>
> The approach advocated by Heikki and yourself gives us 94% availability.
> IMHO that is ridiculous, and I will not accept that as the *only* way
> forwards, for that reason, whoever advocates it or for how long they
> keep arguing. I do accept that some wish that as an option.
No-one is suggesting that to be the only option.
The "wait-for-all-to-ack" looks a lot less ridiculous if you also
configure a timeout and don't wait for disconnected standbys. I'm not
sure what the point of such a timeout in general is, but people have
requested that. Also, setting synchronous_standbys="room1, room2"
doesn't necessarily mean that you have just two standby servers, room1
and room2 might both represent a group of servers.
I believe we all agree that there's different use cases that require
different setups. Both "first-past-the-post" and "wait-for-all-to-ack"
have their uses. There's no point in arguing over which is better.
-- Heikki Linnakangas EnterpriseDB http://www.enterprisedb.com
