[OSM-dev] Data corruption :) II

[Tom Hughes]

Stefan de Konink wrote:
> Tom Hughes wrote:
>> Stefan de Konink wrote:
>>
>>>> Occasional referential integrity problems = 1 (you :-)
>>>
>>> Fix = 1000 + 1 users! And no referential problems anymore + a nice 
>>> reverse proxy infrastructure :) + Probably faster lookups.
>>
>> + synchronisation issues
> 
> Minutely diffs work perfect. The NL crew is already creating a minutely 
> diff tile cache invalidator; where statitics were already implemented.

That would be a backward step - we have perfect synchronisation at the 
moment so why move to a situation where you may get old data, even if 
only a few minutes old?

>> + loss of control unless all proxy targets are maintained by
>>   the OSM admin team
> 
> The OSM admin team maintains the proxy. And targets get checked if they 
> have the latest minutely's every hour -> if not, remove from pool.

Absolutely not. That way leads to chaos. We know this from the t at h proxy 
experience where there was always at least one server that didn't seem 
to be working.

>>>> Seriously, there aren't that many problems considering. It's clearly
>>>> not ideal, hence why it's being fixed. But the fix isn't as trivial as
>>>> you make it sound, so it's been agreed to fix a few other problems and
>>>> introduce some extra useful features at the same time, thus minimising
>>>> down-time for everybody.
>>>
>>> What is the down time you expect for 0.6 migration?
>>
>> We are still attempting to quantify that.
> 
> New server an option?

Moving to a new database server is part of the migration plan.

Tom

-- 
Tom Hughes (tom at compton.nu)
http://www.compton.nu/