Disband unavailable across all services

Timeline

1. Investigating23 Oct 2026, 22:04 UTC

   Disband suffered a major DDoS attack at 5:50 pm EST. All of our inference servers suffered from this attack, and we're working on bringing them all back online.

2. Investigating23 Oct 2026, 22:06 UTC

   We are now learning that the attack started at 5 pm right under our noses.

3. In progress23 Oct 2026, 22:11 UTC

   Our database restarted at 21:59:30 UTC, and Sentinel automatically applied an archive to maintain user data.

4. Identified23 Oct 2026, 22:14 UTC

   We are learning that new data was added to our database and site, which is believed to be malicious. We are actively searching for the data to air-gap it and prune it for later analysis.

5. Identified23 Oct 2026, 22:16 UTC

   Update on the cause of the outage: it was not caused by attack traffic. The requests reaching our servers came from the same people and devices that use Disband every day. What happened: an update to the Disband web app released at 12:29 pm ET today made open conversations check for new messages ten times a second, about thirty times more database requests than before. As more people came online the database fell behind, and from 4:50 pm ET (20:50 UTC) it stopped responding entirely, so every request to Disband timed out. We restarted the database at 5:58 pm ET and service was back by 6:00 pm ET (22:00 UTC), roughly 70 minutes of downtime. Messages sent during the outage did not go through and may need to be sent again. Next: we are preparing a fix that returns message checking to a normal rate, and we are watching the database closely until it is live. Our review of the suspicious data mentioned above is continuing separately.

6. Monitoring23 Oct 2026, 23:33 UTC

   A fix is live. The web app now checks for missed messages once a second instead of ten times a second, and no longer speeds up when its live connection drops, which is what tipped the database over earlier. As people's browsers picked up the update, database load from message checks fell by about 85%, with the same number of people online. Every service is responding normally. We're continuing to watch it.

7. Resolved24 Oct 2026, 01:52 UTC

   Resolved. Disband has been running normally for over two hours since the fix went out, with database load back to its usual level. Summary: an update released at 12:29 pm ET made the web app check for new messages far too often. As more people came online the database was overwhelmed and stopped responding from 4:50 pm to 6:00 pm ET (about 70 minutes). We restarted the database to restore service, then shipped a fix that returns message checking to once a second and stops it speeding up when a connection drops. Sorry for the disruption. We've added a note to that code so this can't quietly happen again.