the error page says :

500: Internal Server Error

A fatal error has occurred within XTools.

The server said: Catchable Fatal Error: Argument 1 passed to MediaWiki\OAuthClient\Client::complete() must be an instance of MediaWiki\OAuthClient\Token, null given, called in /var/www/src/AppBundle/Controller/DefaultController.php on line 128 and defined

Please feel free to report this error on Phabricator (requires a Wikimedia account).

I'm also experiencing the same error with a new user account.

In T183865#3866186, @Matthewrbowker wrote:

I'll take a look.

Thank you :)

Okay, looking through the code, it appears that "oauth_request_token" is not getting set on callback.

@Samwilson @MusikAnimal Who implemented the oauth code? I don't believe it was me...

The xtools-prod01.xtools.eqiad.wmflabs was running out of disk space for storing the session information. I've cleared the cache, and logging in seems to be working again.

The long-term fix for this is, I guess, to figure out what's going wrong with switching to Redis as the cache, or get more disk space and clear it more often. And set up better low-disk space reporting. :)

Also, I was confused initially, because df -h said there was plenty of space on /:

samwilson@xtools-prod01:/var/www$ df -h
Filesystem      Size  Used Avail Use% Mounted on
udev             10M     0   10M   0% /dev
tmpfs           1.6G  155M  1.5G  10% /run
/dev/vda3        19G  9.7G  8.0G  56% /
tmpfs           4.0G     0  4.0G   0% /dev/shm
tmpfs           5.0M     0  5.0M   0% /run/lock
tmpfs           4.0G     0  4.0G   0% /sys/fs/cgroup

Now it says there's 11G free.

I didn't understand anything about the code, but wow thank you guys it works!

In T183865#3866233, @Samwilson wrote:

The xtools-prod01.xtools.eqiad.wmflabs was running out of disk space for storing the session information. I've cleared the cache, and logging in seems to be working again.

The long-term fix for this is, I guess, to figure out what's going wrong with switching to Redis as the cache, or get more disk space and clear it more often.

I wonder if maybe the cache isn't being cleared from disk after it has expired? Most operations are only cached for 10 minutes. I love redis, but even between the two app servers, a shared cache won't be very beneficial to us. With every use of a tool there are several cache reads/writes, so I worry what effect those round-trips to Redis will have on performance (probably trivial?). I'm perfectly ok with giving it a try, but if the current issue indeed is that stale cached data isn't being cleared, we might focus on that first.

And set up better low-disk space reporting. :)

Yes! Adding some sort of monitoring service seems like a good idea, perhaps one that could comment in the IRC channel.