User avatar
🇨🇦Samuel Proulx🇨🇦 @fastfinge@interfree.ca
1w
Is our long international nightmare finally over? Is my server back up after scheduled maintenance? If your reading this, yes it is! All of our collective problems are solved!
2
0
4
0
User avatar
James H @quanin@allovertheplace.ca
6d
@fastfinge My Sidekiq queue thanks you. :P
1
0
1
0
User avatar
🇨🇦Samuel Proulx🇨🇦 @fastfinge@interfree.ca
6d
@quanin But my poor activity pub processor that's now trying to catch up with 60 billion posts from everyone everywhere all at once does not. LOL
1
0
1
0
User avatar
James H @quanin@allovertheplace.ca
6d
@fastfinge As an admin on two of the instances now sending you shit, sorry not sorry. LOL
2
0
1
0
User avatar
🇨🇦Samuel Proulx🇨🇦 @fastfinge@interfree.ca
6d
@quanin Load is only 4.6. I thought it'd be way worse. It was up to 18.4 and exhausted the connections pool for the first ten minutes or so.
1
0
1
0
User avatar
James H @quanin@allovertheplace.ca
6d
@fastfinge Are you using Sidekiq as your activity-pub processor? Because you can tell it to knock that off...
1
0
1
0
User avatar
🇨🇦Samuel Proulx🇨🇦 @fastfinge@interfree.ca
6d
@quanin Nope. I run iceshrimp (iceshrimp.dev). It has its own, capped at 250000 connections.
1
0
0
0
User avatar
James H @quanin@allovertheplace.ca
6d
@fastfinge Oof. Sounds like you should look into ways to make that be kinder to your database. Because yeah, otherwise it's gonna hurt. Sidekiq will stack processes in the queue until a connection frees up.
1
0
0
0
User avatar
🇨🇦Samuel Proulx🇨🇦 @fastfinge@interfree.ca
6d
@quanin Oh, no, the postgresql database is fine. Connections are capped at 200 and reused. But in order to clear queues faster, we accept posts from everyone and just hold on to them until a database connection frees up, so we're not the bottleneck and everyone can clear their queues quicker.
1
0
0
0

User avatar
James H @quanin@allovertheplace.ca
6d
@fastfinge Aha. I might have misread something you said, because it sounded a lot like it was hammering your DB and that was hammering your system load.
1
0
0
0
User avatar
🇨🇦Samuel Proulx🇨🇦 @fastfinge@interfree.ca
6d
@quanin No, it was hammering the number of open TCPIP connections allowed by Linux. And that, thus, hammers load when you have hundreds of thousands of open sockets.
0
0
0
0