The Fix Is One Line in /etc. Just SSH In and Edit It?
Prod is degraded by one wrong sysctl value in /etc on an Ansible-managed fleet. A teammate wants to hand-edit each box now and codify it later. Defend the call.
LinuxYour decision
Prod is degraded because of one wrong kernel/sysctl value in /etc, and a teammate says just SSH into the three boxes and fix the file by hand now, we'll put it in Ansible later. What do you do, and why?
The situation
A misconfigured value in /etc is causing dropped connections in production. It is something like a too-low net.core.somaxconn or an open-file limit that is starving the service under load, and the fix is genuinely a single line. A teammate wants to SSH into each of the three boxes, hand-edit the file, apply the value, and reconcile the config repo whenever the fire is out.
Context
The fleet is managed by Ansible. The value that needs changing lives in a file under /etc that the config source of truth owns. Nodes get replaced during deploys and can be added by autoscaling, and the Ansible play runs on a schedule as well as on demand. A teammate in chat says: just fix the file by hand on all three boxes now, we'll put it in Ansible later.