I really really need some help! β οΈ
For the past two days I have been trying to catch up with the mstdn.social timeline but it's only getting worse.
The default queue is pilling up more and more and it seems I cannot start any more workers. Also the I don't get why the threads are not all working per proccess..
Push & pull are on external servers since when I try to run default queue on an external media upload to Wasabi and email sending does not work.
Please, anymone?
Ive setup some priority workers now but the backlog almost reached a million..
This is so very frustrating, the main server cannot be upgraded anymore since itβs not compatible Hetzner told me, hm
My next solution isβ¦ Another dedicated CPU upper request limit, when granted Iβll setup a new server(biggest they have) and migrate mstdn there and delete the old one..
Iβm not taking no for an answer and I donβt give upβ₯οΈπ₯°
@mdbraber @stux not sure if you were able to get stuff up and running again (not on an instance you control), but double checked with a book Sidekiq in Practice I recently read: thing with threading in ruby is that these do not really work in parallel. There is a GVL; if it is performing some IO task, however, it can release the lock and allow another thread to work. Hence it is better to have more than use more than one thread. But too many will probably block DB connections / IO pools, and push memory consumption over limits.
See also this post by the author of the aforementioned book, @nateberkopec https://www.speedshop.co/2020/05/11/the-ruby-gvl-and-scaling.html
@ChampersBE @stux yeah I would say you might be IO bound. May want to look at iotop?
Also the database might be hitting some limits? You might be locking so much you are basically single threading into the DB.
Another thing worth considering is doing some traffic limiting. Like straight up block large chunks of the world until stuff slows down. I'm not sure the user experience of 404 will be much worse than the system falling over.
The number of max connected clients and pool size seem way too high.
The default_pool_size should probably be closer to the number of cores you have available for your server, and max client connection some multiple of that.
Likewise threads on your sidekiq queues seem high as well. Unless you've scaled out horizontally across multiple sidekiq workers on separate servers, you're probably getting diminishing returns there as well.
@stux Al is het nog zo donker, het wordt wel weer licht!
*you can do this, i'm sure
I would say as a simple girl more investment in hardware
@stux maybe this Post can help you ?
Im not Sure, but it sound like the Same Problem.
Paul Chambers
ππππ
βοΈ
Diane Aurelia Burton
Ben Ramsey
The Febrile Muse
Flying Yogini 
Kio
Neerlandistiek
Maarten den Braber
maarten brouwers
Dragon π
Art πΊπ¦
Bart Champagne ON6BC 
Micah
Nande
Cyber Yuki
Miriam Kronenberg
Adev
Weird
SkyfaR