On May 2, 2013 3:08 AM, "Marlen Caemmerer" marlen.caemmerer@wikimedia.de wrote:
Hey,
On Wed, 1 May 2013, Ryan Lane wrote:
On Wed, May 1, 2013 at 5:03 PM, Tim Landscheidt <tim@tim-landscheidt.de
wrote:
| [...]
There were never answers to this, so I bring it up here again:
- How were "almost /all/ of the problems the TS have had with replication" "caused by that redundancy and trying to keep it synced"?
Ryan, repeaters are from the root of a program inwhich start the initial
setup.
Thanks for this question :) - I also want to know. From my perspective it does not look like this and even the data
inconsistencies appear when we have no commons copy on a mysql instance.
And: DaB experimented with federated tables for commons too and we
decided to not do this since it does not perform from the start.
Probably nowadays when I planned something new in this area (which does
not seem to make sense for TS) I'd really give Galera a try - http://codership.com/content/using-galera-cluster
- What limitation will the Toolserver have at some point?
As to #2: From what I've been told this has to do with future sharding plans for the databases, and due to a change in how we'll be doing replication. Of course, I've heard this in passing. For answers to both
of
these questions you'll need to talk to binasher and/or notpeter on IRC,
as
they are the ones doing the database work.
Thanks for telling...
Cheers Marlen/nosy
Toolserver-l mailing list (Toolserver-l@lists.wikimedia.org) https://lists.wikimedia.org/mailman/listinfo/toolserver-l Posting guidelines for this list: