We are planning a maintenance window of about 30 minutes on 2020-01-28T22:00:00Z. During that time WaniKani will not be accessible. Our host requires us to do some maintenance on our database.
Updates will be posted when they happen.
Thanks for your understanding and patience.
It seems like they do that a lot. Just curious, can they not update some of the database servers, then update the rest after those are done (switching over the master node as needed)? I’m not super knowledgeable about database stuff, but I figured I’d ask.
Is 6:30 Tokyo time also ok? Nah I’m just kidding but unlucky it’s the time Where I usually Start studying
thanks for the heads up 
May it go well.
Thank you for making the site better!
By the way - what IS your time zone viet?
That’s Greenwich Mean Time, not some random american time zone right ?
宜しくお願いします。|ﾟーﾟ)ﾉ
Yeah it has been more frequent as of late. They give us a time when they plan to do it. But we are given the option to do it on our timeframe as long as its ahead of the date.
How it is done is we spin up a follower database, which has the updates required. The follower replicates the master. Once it is caught up we can detach the follower from the master and then promote it. It is this period where we need to “freeze” the writes to the database. This swap shouldn’t take more than a few minutes. It is during this period where we need the site to be in maintenance mode. Us being cautious we just say 30 minutes to give us time to deal with any issues.
Pacific Timezone.
The way you formatted it in the OP, it shows me my local time, though. When clicking, it shows me both, my time and your time.
very cool oke!
I have no idea why but that time conversion window fascinates me like a child looking at a clock for the first time.
おはようございます！
了解です！
Even if your host doesn’t require it, it’s always a good idea to do periodic maintenance on databases, like vacuuming in postgresql, etc. They’re terrible at garbage collection.
I don’t know which backend you use, but if you do it right, you can failover without having to bring the application down. Requires a little design and planning first, though. I use mongodb at work and you can fail over to any one of the secondaries and then add a new node to the cluster… pretty sure that works.
To be fair, we have no end of sync problems in postgres. I guess some things are just intractable.
With a master / slave configuration (there is such a thing as a master / master configuration), that doesn’t necessarily work. I also don’t know much about hosting databases (I’m a developer) but this is not as easy as it sounds and could be super error prone. I’d rather have half an hour downtime.
Is this related to the issues with the subscription not being immediately available after buying it for some users?
Looks like Heroku judging by the Via: 1.1 vegur HTTP response header.
Aaaand here we go! See you on the other side! (and along the way)
