Postmortem Index

Explore incident reports from various companies


An operator used a tool with lax input validation to reboot a small number of servers undergoing maintenance but forgot to type -n and instead rebooted all servers in the datacenter. This caused an outage that lasted 2.5 hours, rebooted all customer instances, put tremendous load on DHCP/TFTP PXE boot systems, and left API systems requiring manual intervention. See also Bryan Cantrill’s talk.

Source | Edit Metadata | JSON file