I am currently getting signed out every minute from lemmy.world. This is not a client side cache issue. I tested making API calls from the command line (with curl
) with no cache and the issue still occurs. One call I get the correct response, the next I get a 400 telling me im not signed in.
I’m primarily testing with the https://lemmy.world/api/v3/user/unread_count api endpoint. I’m not sure if this issue occurs with all endpoints.
Reproduction steps:
- Get a lemmy.world JWT token for your account using your desired method (eg. postman).
curl https://lemmy.world/api/v3/user/unread_count?auth={JWT_TOKEN_HERE}
- Note the 400 error. If you do not get an error repeat step 2.
Edit
This issue only seems to affect lemmy.world so a temporary workaround is to use a different instance for the time being.
Just a quick statement from the admins team to say that we are aware of the issue and yes we are looking into this.
Thank you @idunnololz@lemmy.world for the elaborate report and everyone else for their patience while we try to sort this one out!
Edit: Lemmy was upgraded to 0.18.2
Thank you for all that you do for this place. I am consistently amazed at how quickly y’all are able to resolve issues.
o7
Thank you for making a statement about it!
Sounds like lemmy.world runs on 2 instances and the requests are being loadbalanced between those two. That and that the jwt secret is different between those two instances causing one to accept and the other to reject
This is also my theory. I think you’re right on the money here. They probably rotated secrets from yesterday’s hack and forgot to restart both servers.
Does anyone know who can contact the server admins?
Yell real loud in all caps
REAL LOUD
Same problem for me it seems, dunno if I’ll even be able to comment. Refuses to stay logged in.
From my tests, it’s almost perfectly a 50/50 whether any API requests you make will yield a 200 (success) or a 400 (not signed in). If you perform an action that takes 3 API requests, your chances of succeeding is (1/2)^3 or 1/8 because only 1 request needs to fail in the chain for the entire action to fail. So, as long as you make single API actions you can maximize your success rate :D
Seems like spamming actions also gets it to work eventually. It’s a pain in the arse though lol. I made some alt accounts on other instances, but I’m lazy and don’t wanna rebuild my subscription feed if I don’t have to, so hopefully it gets fixed at some point.
What’s an example of something that would take more than one API request?
Signing in. Most websites/apps will probably also grab your unread count, and maybe even your subscription feeds.
Another example is checking your inbox. Lemmy actually has 3 inboxes: mentions, replies and PMs. A lot of websites/apps bundle these three so they will need to check all 3 inboxes via 3 API calls.
Smells like two instances behind the load balancer, one is fine with the JWT, one is not.
Thaaats what’s going on
Yeah. Lemmy.world is currently unusable on the desktop. I don’t have that problem in Memmy. Growing pains but I hope the problem will be fixed soon. Do anyone know if one of the mods in North America are aware of the problem?
I was having trouble in liftoff and the browser. Cleared data and cache from liftoff thinking maybe something got messed up there and now I can’t even log back into my .world account 🤷♂️ I’ll hang here for a bit I guess.
Same here, can’t log in again via Liftoff.
I’m choking in desktop browser and in liftoff. Jerboa seems ok. It’s weird to me how different clients react differently, I’m not sure how they interact differently.
Hey! you actually gave useful information. Some people are just grunting or snapping at people hoping someone figures it out lol.
Same issue here, I’m being automatically logged out of my lemmy.world account in Firefox. If I refresh the page even immediately after logging in, I’m automatically logged out.
I’m having to reauthenticate in safari and wefwef every time I load a new page. Furthermore, the login is frequently failing.
Login in likely always succeeding. The issue is that whatever app/website you use will make additional API calls afterwards (eg. fetch posts or fetch unread count). Each of those calls have a 1-in-2 chance to succeed and if any of them fail, they all fail and you will be booted out.
Lemmy is now an RNG game. We must prayge to rngesus before making any actions.
Schroedinger’s API call.
I’ve been experiencing something similar/related. If I am logged in and open something in a new browser window, it frequently (starting today) shows me as not logged in. If I refresh the page, I’m suddenly logged in. This doesn’t feel like a authentication problem as much as a timing issue while loading the page. Or maybe what I’m seeing is an entirely different issue.
At least when you can’t log in on one instance you can just login on another. Downtime doesn’t mean you have to go do something else anymore!
I’m seeing the same issues on my app, calling login, then immediately using that jwt to fetch the site details and it doesn’t give
my_user
half of the time, and if my app loads far enough to check the unread count I getnot_logged_in
Same here, I thought I was going crazy
I can’t seem to comment on a couple specific posts on the instance. But as you can see, it works on this one. I am wondering if that’s related? I’m not even on my Lemmy.World account and get an unable to post error as soon as I hit the button, like it’s not even trying to do anything.
From my experience it’s entirely random. You can make 5 actions and all 5 will work. Then have a string of 5 actions where none would work.
Yes I have the issue on Liftoff - can’t even log in again. Alt time for now!
Same here, it’s driving me mad. Also did the above and glad to see I’m not alone!
The good news is it only appears to affect lemmy.world. If you have an account on another instance, you should switch to that account for now.
I might do that at this point, with the attention world is getting, it might be smart to have a backup.
Just add me to the list. Jerboa seems to stay logged in for about 75% of interactions.
Okay, so how do we get this fixed? Any way to get admin attention? I think @Spaltovic@lemmy.world is probably correct about the cause.