There is a bug that in a very specific situation, instances running version 2.3.x (not sure if others) can get a stuck Sidekiq request to /users/username/collections/featured
This causes an endless loop of that request and could bring my servers to an halt. So, I had to make the decision to block several instances that were causing a massive amount of load to the servers in the shared infrastructure.
Currently the blocked instances are: pentacl.es cybrewitch.es social.infranix.eu mstdn.dolphinbox.net
I couldn’t find a way to contact some of the admins from those instances (plus via Mastodon they are blocked on my end) and I cannot unblock them unless their instances are updated to at least v2.4.0 and their Sidekiq queues cleared from this request.
If you know any of the admins from those instances and can reach to contact with them, ask them to update their instance to at least version 2.4.0 of Mastodon and to clear their Sidekiq queues or any instance hosted in Masto.host shared infrastructure will not be able to get federation from them.
If the instances in question cannot upgrade, they can apply this hotfix (and clearing their Sidekiq pull queue) that should fix the issue.
Plus, they should have their Sidekiq queues filled with failed requests and their instances are using way more resources than needed or simply have their federation stuck.
Here is a partial screen grab of the request from one of those instances to one of Masto.host Host servers for a couple of seconds:
If any of those admins needs my help with this, just contact me info@masto.host and I will do my best to help.
This was a tricky decision as I don’t want to block any external instances and there were some doubts in my mind if I should do it but I decided to go ahead because that was the only way I could maintain all instances hosted with me running without problems. If you have any suggestions on how to handle situations like this better, please let me know @mastohost@mastodon.social or via email.