Jerry 🦙💝🦙Nov 27, 2022

I just wrote a post about the current state of growth in infosec.exchange at the one month mark post E-day here: https://blog.infosec.exchange/2022/11/27/an-update-on-growth-of-infosec-exchange/

Note: I installed a plugin that will allow you to follow blog posts there by following @[email protected]

An update on growth of infosec.exchange – Infosec Exchange Blog

Show threadRoy ThomasNov 27, 2022
@jerry @[email protected] Thats over 170X increase in users in 30 days :) I used to be a Sysadmin for a long time, and was responsible for multiple data centers so I know the challenges. But no one sane enough builds a just-in-case plan to increase capacity 170X within 30 days :)
Show threadJerry 🦙💝🦙
@chromedevice @[email protected] it was an exciting time in my life, to be sure. We grew from one server to 10 in that period. The 10 gives a bunch of headroom, though.
Nov 27, 2022 at 4:44pmWeb
Show threadRoy ThomasNov 27, 2022
@jerry @[email protected] Do you have a recommendation on where more architectural discussions are going on for Mastodon scale challenges ? For example: if you are running 10 independent servers, is someone working on optmizing Mastodon to reduce egress traffic (or URL pings to servers) and allow your 10 instances to share resources ?
Show threadJerry 🦙💝🦙Nov 27, 2022

@chromedevice @[email protected] that's a good question. The 10 servers comprise the single instance that is infosec.exchange.

There are two front end web servers running nginx and puma.
There are two sidekiq app servers that only run sidekiq jobs
there is a single postgres database server
there is a single cache/search server
there is a single (moving to HA) pair of minio servers that will be behind a load balancer
finally, there is a management server for orchestration, monitoring, and so on.

Show threadRoy ThomasNov 27, 2022
@jerry @[email protected] Got it. So this is still collectively one instance which is scaling vertically from the database point of view.
Show threadJerry 🦙💝🦙Nov 27, 2022
@chromedevice @[email protected] more or less, yes. The database is, as it turns out, the least busy part of the entire operation