Hi, I started my discourse forum on a digital ocean droplet with 4 gb ram 2 vcpu 60 gb ssd.
My site grow and in certain moments users got the message “due to extreme load, the site is shown as a logged out user”. It’s a sport site and it happened on some events with a lot of interest.
I think the message appeared when we reached 250 connected users.
in App.yml I have 2 unicorns and 768 MB db_shared_buffer
I want to resize my droplet, which size do you suggest?
Thank you
AI-generated summary
The discussion revolves around optimizing a Digital Ocean Droplet for a Discourse forum. Saiano experiences overload issues with 250 connected users on a 4 GB RAM, 2 vCPU, 60 GB SSD droplet. sam suggests increasing the number of web workers, recommending 8 unicorn workers for a 4 CPU, 8 GB RAM setup. sam also advises allocating 1 GB RAM to Postgres. To implement these changes, codinghorror recommends re-running ./discourse-setup to adjust the app.yml settings based on the detected memory and CPUs.
You may want to wait as some bad code was checked in that added multiple N+1 and N+2 queries cc @sam will latest have the fixes yet?
Latest has the latest fixes, but as a sport site you may simply be overloaded.
2 unicorns does not sound like enough, for 250 active non anon users you would need more web workers.
Thank you, which size do you suggest? Is it enough 8 gb ram and 4 cpu? Unicorn numbers will auto update rebuilding app?
if you have 4 cpus and 8gb, you can probably get away with 8 unicorn workers, also be sure to give 1gb or ram to postgres.
how I can check this setting?
Re run ./discourse-setup and it will make the app.yml adjustments based on memory and CPUs detected.