The following primary database masters will be switched over during the
next few weeks (more details at https://phabricator.wikimedia.org/T230788):
*Writes will be blocked*
*Reads will remain unaffected*
These are the concrete days, hours and affected wikis:
* s8: 10th Sept from 05:00-05:30 UTC. The affected wiki is: wikidatawiki -
tracking task: T230762
* s2: 17th Sept from 05:00-05:30 UTC. The list of affected wikis is at:
- tracking task: T230785
* s3: 24th Sept from 05:00-05:30 UTC. The list of affected wikis is at:
- tracking task: T230783
* s4: 26th Sept from 05:00-05:30 UTC. The affected wiki is: commonswiki -
tracking task: T230784
If everything goes well, we do not expect to use those 30 minutes of
read-only and rather just a few minutes.
We will email send an email the day of each failover before and after it is
Sorry for any inconvenience this might cause.
The current primary master for m1 (db1063), which is mostly for internal
services + etherpad isn't in a great healthy status: it is an old host,
which needs to be decommissioned and which is having disks failing pretty
much every week (plus disks on predictive failure).
We have decided to fail it over one of the newer hosts, db1135:
We have scheduled this switchover for: Tuesday 10th September at 16:00 UTC
This failover should be rather quick and would only take a few seconds
(while we re-load the haproxy) - during those few seconds, the following
services will be on read-only:
Communication will happen at #wikimedia-operations
If you are around at that time and want to help with the monitoring, please
It's time for Wikimedia Tech Talks 2019 Episode 7! This talk will take
place *Sept 4, 2019 at 6PM UTC*.
*Topic:* Documenting Wikimedia technical projects
*Speaker:* Sarah R. Rodlund, Technical Writer -- Developer Advocacy
This talk will discuss what technical writers do and why they are critical
members of the Wikimedia technical community. You will learn more about the
skills needed to be a technical writer and how to build these skills by
participating on Wikimedia and other open source projects.
The talk will also cover some ongoing initiatives to improve technical
documentation for Wikimedia projects.
*YouTube stream for viewers:* https://www.youtube.com/watch?v=GKvyk4VAPb0
During the live talk, you are invited to join the discussion on IRC at
You can watch past Tech Talks here:
If you are interested in giving your own tech talk, you can learn more
Sarah R. Rodlund
Technical Writer, Developer Advocacy
We are going to switchover m5 (wikitech) primary database master on *Tuesday
3rd September at 13:00 UTC *https://phabricator.wikimedia.org/T229657
This switchover will also affect some Cloud services that will be handled
on a different announcement.
The current master (db1073) is an old and out of warranty hosts with
several disks on predictive failure, it needs to be decommissioned.
We will be failing over db1073 to db1133, a newer and more powerful host.
We expect this to take around 15 minutes (or just a few seconds if
everything goes fine).
*- Writes will be blocked during the switchover*
*- Reads will remain unaffected.*
Communication will happen on #wikimedia-operations IRC channel.
I see that in some classes, like WANObjectCache, most methods are declared
final. Why is this? Is it an attempt to optimize?
The problem is that PHPUnit mocks can't touch final methods. Any ->method()
calls that try to do anything to them silently do nothing. This makes
writing tests harder.
If we really want these methods to be marked final for some reason, the
workaround for PHP is to make an interface that has all the desired
methods, have the class implement the interface, and make type hints all
refer to the interface instead of the class. But if there's no good reason
to declare the methods final to begin with, it's simplest to just drop it.
Hey all, the Timeless grant has wrapped up and I've submitted the final
report for the project. Please check it out if you're interested:
A couple of things to highlight:
* The associated learning pattern is likely going to be relevant to a lot
* Part of the purpose of the grant was to investigate into what we would
actually need from a new skin on Wikimedia projects, should anyone
decide to go that route. Here's what we found:
> All we really need from a new skin is for everything that went wrong
> with Minerva and Timeless, as far as readers and editors are concerned,
> to not happen this time:
> * We need gadgets to work.
> * We need content to display properly, and tools to be available.
> * We need to properly reuse core interfaces so things remain consistent.
> * We need to not distract users with unnecessary colours and affordances
> that aren't relevant to whatever they're doing, and provide options
> that suit their differing situations and needs.
> * We need to not do dumb things, like reinvent random wheels or ignore
> standard hooks and DOM expectations.
I don't know if that's all that likely to help anyone, but... yeah.
Anyway, I'll still be working on maintenance and some feature development
in a volunteer capacity, mind you, so as always if you run into issues or
whatever, please file them on phabricator or come yell at me! I just make
absolutely no promises whatsoever as to when I (or whoever) will actually
deal with it, but it's still really helpful to at least have them tracked.
(See, it's even sort of organised! I even renamed the 'Aaaaaaa' column!)