#mastoadmins
It's time for a new update about the #tootcommunity finances. 💰
Spoiler alert: We're no longer operating at a loss. 🙏
https://ko-fi.com/post/Server-costs-per-November-25th-2022-N4N1GLN9H
New #MastoAdmins are running on adrenaline, and I suspect they'll start getting tired very soon. We'll possibly begin seeing new and fastly grown instances announcing their early retirement soon, and I hope we'll grow despite the pain and level up our commitment. Still, I worry that not all will and that the #fediverse will have to endure short-term pain to be better off long-term. 4/5
What if a network interface breaks down or a small percentage of network traffic turns wrong? New and recently born #MastoAdmins with significant user counts converge at cheap places since they're paying it out of their pocket. Technical support will be slow, and I worry that from a user perspective, this could damage the image of decentralized social media if user-facing issues take too long to resolve. 2/5
There's been some talk about how big a Mastodon instance should grow. Ultimately it's up to the #mastoadmin to push the buttons, but from where I stand, one should not forget topics such as accountability, liability, and (financial) stability. We're helping people grow their social network and supporting businesses in reaching their new potential clients, and I feel that we're shrugging it off. 1/5
Sorry guys, we're limiting user registrations again. Servers are still scaling & very happy but need to consider the long-term cost to keep this thing sustainable.
Its been a long day and I'm going to spend some time with the family. Please don't break stuff while I'm gone 🙏
#devops statistics of #tootcommunity for the last 6 hours:
- 4.33 million requests (+61.1% increase)
- 151.9 GB transferred (+64.5% increase)
- 53.8k page views (+81.6% increase)
Servers right now: :partyparrot:
#devops statistics of #tootcommunity for the last 6 hours:
- 2.45 million requests (+50.6% increase)
- 86.03 GB transferred (+22.9% increase)
- 27.5k page views (+57.1% increase)
Noticed occasional spikes in latency for #tootcommunity. We were just setting `requests` on sidekiq runners to help scheduling, but don't have limits. When the default queue started filling, the sidekiq workers took all the available CPU cycles resulting in poor performance for the end-users. Solution: Split up the node pool, add taints/tolerations so only the sidekiq processes run on those machines.