Post

Dreams do come true. @selfhosted.social is now running on two PDS containers to spread the work across multiple node processes. Bad news. It did not help with CPU usage being pegged at 70%. It is now 100%. I guess if you're going pay for something, use it all.

4 reposts74 likes

TLDR: - redis for shared state between the two - caddy for load balancing. just doing round-robin - need to set new env variables including PDS_DPOP_SECRET (I can never escape oauth) Cool thing is it's all built in and works on the ref PDS

Have had a couple of theories - Had a faulty blob that was returning errors and 90+ requests every 5 mins. blocked it at caddy, no help - I did find someone walking the repos manually for a very extended time. blovked - New working theory is tap. There's almost always a getRepo going. all diff ips