Het optimaliseren van de droplet na een upgrade van de servercapaciteit

AI-generated summary

ckshen upgraded their server’s RAM from 2GB to 4GB to address frequent ‘extreme load’ messages, while keeping CPU cores at 2. Post-upgrade, disk I/O improved significantly, but load averages remained largely unchanged — with anecdotal reports of more frequent extreme load messages and slightly more frequent load spikes over 1.0. Memory usage dropped from ~80% to 60%, and disk usage was reduced from 86% to 66% after cleanup. ckshen is using 4 Unicorn workers, which they believe is appropriate for 4GB RAM, and is asking if other settings should be adjusted. pfaffman suggests re-running discourse-setup to auto-adjust memory settings or manually editing the yml file, which includes helpful comments explaining each setting. Graphs provided show system metrics before and after the upgrade, with no obvious anomalies beyond the load behavior. The discussion centers on optimizing server performance post-upgrade, particularly around load spikes and worker configuration, under the context of self-hosting support and server resource management.

Hi,

We upgraded our server capacity on 11/24 by doubling the RAM from 2gb to 4gb, while keeping the CPU cores at 2. We upgraded because users have been complaining about seeing the ‘extreme load’ message often. On the whole, as you can see from the graph below, disk I/O reduced significantly after the upgrade. While load seems to largely stay the same.

This is anecdotal of course, but some users say they are seeing the Extreme load message even more frequently than before. I look at the load graph again, and maybe the load spikes over 1.0 a teeny bit more than before.

I read here before that a recommendation that Unicorn workers should be proportional to the amount of server memory. Our set up is currently at 4. I thought that’s in line for the amount of memory that we have. Are there other settings that I should be tweaking?

Thanks!

Some other graphs that I didn’t see anything out of the ordinary. Memory utilization steady at 60%. (at ~80% before the upgrade)

Disk usage at 86%. If that matters, I just did a launcher cleanup and reduced it to 66%.

If you run discourse-setup again it’ll adjust the memory settings. You can also edit the yml file by hand. There are comments that describe the settings.