Without seeing any code my Google searches are as good as anyone else’s but I found this Which seems to cover most cases of those error happening.

Seems to me like too many writes from too many processes. (More than 100 specifically if the defaults haven’t changed since 2008)

You can (and probably should) use sync transactions as mentioned in the linked blog post. And it may be worth playing around with the dump threshold settings but in general the overload event means your system is doing more than it is capable of. If these events keep coming you need to have more resources. If the overload is just temporary it is fine. (I.e a short burst of writes coming in).

That said, one thing that helps with performance is to limit the number of concurrent writers. In my experience having 1 writer per scheduler seems to work pretty well. I think I picked this up from a talk from WhatsApp but can’t remember

You can also subscribe to the overloaded event and try to be proactive by using a circuit breaker of some sort (GitHub - jlouis/fuse: A Circuit Breaker for Erlang · GitHub for example). If you run mnesia in a distributed setting this may be a bit harder.

One talk that that briefly mentions this is Mnesia for the CAPper (Erlang Factory - Ulf Wiger, Uber Erlang programmer and CTO of Erlang Solutions) (I’m sure there is a video somewhere as well). This talk also contains a few other things that is good to know about mnesia.