Skip to content

Comment on Post mortem on Mastodon outage with 30k usersparent

Comments

Hi. I wrote the post. Additionally I am responsible for operating Hachyderm (Ruby on Rails) and GitHub (Ruby on rails) for both my free-time and my day job.

I can say with certainty that Ruby specifically was not the bottleneck in our case. I do think that the rails paradigm can often lead to interdependent systems. We see this at GitHub and we also see this in Mastodon. Service A will do reads/writes against the same tables in the database that Service B also does. When service A is moved to an isolation zone, it can still impact Service B's performance.

In other words, I think any stateful framework with the flexibility that Ruby on Rails encourages bad behavior that can contribute to a noisy neighbor problem.

The point I am trying to drive home is that I agree. I can confidently say that Ruby on Rails is not the culprit in our case. To be honest I just ignore anyone who is quick to point fingers and assign blame either technically or personally.

Sorry hacker news got you down. If it helps my family and I are making Sunday morning pancakes with my puppy Björn today and we are all wishing you the best day ever.

aw thanks! I follow you on mastodon and it’s been great watching the updates as hachyderm grows

Is there any good architectural breakdown of Mastodon software?

AboutSource Built by g1lg1l

Hackerly is an independent reader for Hacker News, built on the public HN API. Not affiliated with Y Combinator.