Χίμαιρα corvids @autistikai@computerfairi.es

Inferences I'm taking from what I can see about the Discord/GCP outage:

1.) Discord's database and storage is effectively single homed. If they had a hot failover it woulda failed over. If they had a cold failover its cost to fail over is probably high or else they would have.

2.) Discord's DB is intolerant of single-region networking issues. Google reports 8% packet loss in one cluster in one DC, and that knocked Discord offline.

3.) Discord may or may not have backups (they better) but the restoration time for a dataset of their size would be days since it's single-homed.

4.) Discord is reliant on Google Cloud. If they had their infrastructure striped across multiple providers, see point 1.

5.) Discord has no response process for if Google Cloud just completely soils the bed, as noted by their updates saying "All we can do is wait for GCP to fix itself"

6.) Discord has no way to gracefully handle a failure between API service nodes and their monolith DB/cache. Granted the nature of the app limits what could be done, but "Wait and hope the message doesn't redline" isn't good and just increases load in a fragile situation.

Synthesis.) A major incident could knock Discord offline for days, or permanently, depending on the reliability of their backups. Absolutely have backup methods for reaching your important contacts.