[Labs-l] [Toolserver-l] TS and/vs. Labs

Mon Sep 16 23:13:15 UTC 2013

On 17/09/13 00:11, Tim Landscheidt wrote:
> Ryan Lane<rlane at wikimedia.org>  wrote:
>
>> [...]
>
>>>> I'm more than happy to recommend a number of cloud services and am
>>>> more than willing to give advice on how to configure and run tools
>>>> and bots from those services. It's even possible to reuse the work
>>>> we're doing in the tools project, or in the Wikimedia infrastructure
>>>> via our puppet repository since our infrastructure is Open Source.
>
>>> Very nice idea – how I get the mysql-replication-stream? I got several
>>> offers of donation if the Toolserver would continue; the only problem is
>>> the replication-data. But because the data is open-source, it shouldn’t
>>> be a problem than, should it?
>
>> Assuming you found a non-profit, host your infrastructure somewhere that
>> doesn't cause legal issues and every person that has access to the data
>> stream signs an NDA it's likely doable. [...]
>
> The NDA isn't necessary.  According to
> https://wikitech.wikimedia.org/wiki/Tool_Labs/Database_plan,
> the data set at the LabsDB stage is free of non-public data
> (modulo MariaDB accounts information which should probably
> not go off-site even with an NDA :-)).
>
> So we could (and IMHO should) provide DB dumps/bin logs at
> dumps.wikimedia.org or somewhere similar to anyone who can
> download them.
>
> Tim

labs db server do contain the non-public data. It's just not viewable. 
So there aren't bin logs for just the non-public data.
You *could* make a dump of the database (probably creating a new tool, 
as mysqldump would simply dump the view definition)... assuming you that 
in doing that you don't kill labs filesystem. ;)