Skip to content

Emergency: start here

Current incident: Drupal core security release on 2026-11-18, 17:00-21:00 UTC

Critical Drupal core release for all supported branches (PSA-2026-11-16), 18:00-22:00 CET. Prepare before the window: Drupal security release.

Every host in the Ansible group traefik of a client that uses the client template has the tool lakedrops-emergency. A measure takes effect in seconds: no restart, no commit, no image build. The host's state decides, and only lift removes a measure: neither the nightly site run nor any other run of the role, nor the date of --expires.

This section tells you what to do. Every detail is in the toolkit's reference runbook, docs/emergency.md.

What is happening?

Emergency Measure
Vulnerable app, no patch yet: it must go offline, with a notice shutdown
One site is attacked or may be compromised shutdown --host or --stack, then Suspected compromise
The exploit shows in the path, query, method or headers block
The exploit shows only in the request body waf rule
The CrowdSec hub publishes a virtual patch for the CVE waf update
A mitigation inside the apps: uninstall a module, revoke a permission, change a setting exec
A flood of requests throttle
A botnet whose requests look like visitors (not for Nextcloud) challenge
Known attacking addresses or ranges deny
Known addresses that must be cut at once, open connections too deny --nft
A flood that fills the uplink Nothing on the host helps: the provider's protection or a CDN, see Floods that fill the uplink

Path or body? See block: path or body?.

Where do I run it?

Scope How
One host SSH over NetBird, sudo lakedrops-emergency …
One client Run pipeline with ASCR_COMMAND = emergency …, ChatOps run ascr emergency …, or ascr emergency …
All clients lda-fleet -- … on a workstation with Docker

Prerequisites, examples and the hosts that are not reached: Where to run a command.

Golden rules

  1. Preview first. On the host --dry-run; with ascr and lda-fleet --preview (the tool's --dry-run on every host). lda-fleet --dry-run only lists the clients and the command, it starts nothing.
  2. One incident id per measure that you may want to lift on its own. Ids have 1 to 40 lower-case letters, digits and dashes, and start and end with a letter or digit. A measure with the type and id of an active one replaces it; lift --id removes every measure of the id.
  3. Check after acting: status, and a request from outside (each measure on Measures says how).
  4. Never edit by hand /opt/traefik/configuration/emergency.yml, the notice pages or the tool's AppSec files: the next render writes them again.
  5. lift when done. While a measure is active, ascr sanity check keeps the issue [<host>] Emergency measures open; it closes after the last lift.