jae

🛠️ If you haven't yet, please read the introduction and check out the bug tracker to see what we’re working on. Now onto the patch notes!

a day late! we landed exactly zero user-facing features, but we fixed lots of bugs. friday was, internally, the first installment of what we’re calling Bug Fix Friday; you can probably guess what that is. lots of bug fixes and internal improvements this week.

bug fixes

• fixed a bug where you couldn’t follow anyone because the button was invisible
• fixed a bug where liked posts wouldn’t show up across reloads
• fixed a bug where notifications from pages you follow but have silenced would show up in your feed
• fixed a bug where deleted comments would still show up in your notification feed
• fixed a bug where typing in the post editor would fully re-render the page, causing gif playback to restart (among other things)

everything else

• strict null checks landed! check out volume 13 for details
• changed how data moves around internally for some pages, part 1 of a refactor that will eventually cover more of the site
• lots of little changes to our deploy, testing, and performance metric collection processes
  • turns out node >16.10 has a serious memory leak with our test runner! had to roll back so tests could ever finish
• improved our internal maintenance tooling so we can fix live data issues more easily

outage follow-up

in addition to Bug Fix Friday, we hope you all enjoyed the first time in a while that cohost has ever been fully Bug Free, insofar as it was completely unreachable: approximately 3:25 to 3:35 yesterday afternoon. the reason we went down is very silly: jae accidentally deleted the deployment from our kubernetes cluster because they thought they were working on their local cluster. whoops!¹

the good news is no user data was lost, and our improvised incident response process worked great; after we discovered that the site had vaporized, we re-ran the last deploy on the servers that run cohost, and the site was available again within 90 seconds.

the bad news is that it happened in the first place; and, after it happened, it took us a few minutes to notice it was gone. in response we:

• (finally; it was overdue) got alerting ironed out, so we get automated notifications if the site is unavailable, instead of the start of our incident response process looking like this:
• are planning to tighten up our development workflow around keeping credentials that can modify the deployed installation of the site on our work machines only when absolutely necessary; right now we’re treating cohost.org as if it’s a development environment, but we need to make the transition to treating it as if it’s a real thing with real stakes soon.

coming up on cohost

• colin is almost done with the moderation tooling; sorry that it’s taking so long, and it’s not showier, but there’s a lot of different small things that need to be cleaned up along with it.
• jae got so stressed out about money that they’re working on our first Revenue Generating Product (ugh): cohost Plus!
  • we’ll have more details about cohost Plus! soon
• aidan is working on our launch marketing materials. we swear it’s actually gonna happen soon.

1. they would also like to point out that they wrote this, no one is throwing them under the bus.