postmort: july 31, 2023 outage
postmort: july 31, 2023 outage
we suffered some extremely unexpected downtime while I deployed a trivial change (a reverse proxy from http://awful.systems/archives
to http://these.awful.systems/archives
) to prod
the downtime was unrelated to the deployment change; instead, it seems like lemmy-ui
started crashing because it couldn't render the app icons it uses when saved as a home screen app on mobile. it uses a fairly heavy dependency to do this, and has no error handling in case the source icon data is corrupt, which causes it to crash on every request (resulting in a 503 Service Unavailable
error for everyone who tried to access awful.systems during this outage)
since I don't know how that corruption occurred or why it was persistent (the app icon data should be fully static as part of the Nix store as far as I know), so until I can dig in I've disabled generating app icons for our instance. since it seems like we're the first ones to hit this bug, I'll do my best to keep the patch upstreamable so other lemmy instances can benefit from the fix