קבלת אזהרת עומס קיצוני בזמן שמשאבי השרת אינם בשימוש

סיכום מבוסס בינה מלאכותית

A game company using Discourse for their community forum experiences an “extreme load warning” during patch note releases, despite low CPU (12%) and RAM (40%) usage on their 16vCPU/32GB DigitalOcean server. Caliph increased capacity ahead of a major patch, but the warning persists — a recurring issue even after upgrading to Discourse beta4. IAmGav suggested reviewing a beta6 bug fix thread, but Caliph doubts it’s relevant. Falco recommended enabling the Prometheus exporter plugin to collect performance metrics for tuning, and noted that at $320/month, their current setup may be overpriced compared to Discourse’s dedicated hosting. Caliph plans to explore metrics and asked if New Relic is supported; Falco clarified that only the Prometheus plugin is officially supported, as the old New Relic integration is unmaintained. Caliph noted they typically run fine on a $40/month 4vCPU/8GB plan handling 1M page views, suggesting the issue may not be raw capacity but rather Discourse’s internal load detection or configuration.

We are a game company and use discourse for our community forum.
Each time we release a patch to our game and patch notes on the forum we get a spike in traffic.
We are using DigitalOcean to host and recently increased server capacity to 16vcpu and 32GB ram for the upcoming patch release (which is tomorrow)

We just released patch notes today and it made the forum show the extreme load warning when I checked DO graphs the CPU usage was only at 12% max and ram at 40%

I have already set the unicorn workers to 32 (double the number of CPU cores)

What am I missing here?

We usually get 100k visits the day of the patch, of which ~30% of users are logged in

You might want to read this

I just updated the forum yesterday to beta5 :cry: I will update again tomorrow.
But I feel it might be unrelated to our issue as it is happening to us for a long time now.

Edit: Just checked I upgraded to beta4 so this won’t be it.

After a certain traffic threshold you need to start collecting performance metrics in order to fine tune your install.

That costs $320 right? At that price point you may want to seriously consider our dedicated hosting service, which is both cheaper and faster for this traffic level.

I will look into performance metrics, thanks for that. A question related to that - Can I use newrelic for performance metrics?

About hosting, as I mentioned we recently did this upgrade for patch release and will revert it back to basic 4cpu 8gb which is just $40 per month and we serve 1m page views without any problem :man_shrugging:

We only officially support the prometheus plugin. There was an old New Relic one but it wasn’t maintained and broke over the years.