When something's wrong — Cheat Sheet

Most "broken" moments aren't broken — they're stale, or a task that simply hasn't finished yet. This page is the quick primer: the first things to try, how to read a task that failed, how to get back into a site you're locked out of, and how to tell what's yours to fix from what's your host's. Every fix here runs from the Ægir control panel or your own shell — no root — and everything links to the full story.

The first things to try, in 30 seconds

Wait about a minute, then hard-reload. A saved change can take ~60 seconds to appear — BOA holds pages ready-made and keeps compiled code hot. → Start here
Flush all caches. The panel task, or from the shell drush @example.com cc all (Drupal 6/7) / vdrush @example.com cr (Drupal 8+). → Site-health tasks
Run Verify. The safe "fix it" button — it re-writes the site's config (nginx vhost, settings.php, aliases, DB connection) to match disk and can't hurt anything. → Task recovery
Check the last task actually finished — status Successful, not Queued, Processing or Failed. A lot of "it didn't work" is really "it hasn't happened yet." → Task recovery

A task failed? Read the log before you retry

Open the site's node → its Tasks list → the failed task, and scroll to the bottom of the log — the real error is in the last few lines, not the top. Fix the cause, then press Retry. A blind retry on the same broken state usually fails the same way, or leaves things slightly worse. (Run Again is the button on a task that already succeeded — not the same thing.)

→ Failed and stuck tasks

A platform stuck in a weird state

If every task on a site suddenly fails, or you see Drush-version errors, the platform underneath needs a nudge. From the platform node, run three tasks in order: Verify + Lock Drush → Unlock Local Drush → Verify + Lock Drush (yes, that last one again).

→ Failed and stuck tasks

White screen? Read the hidden error

A blank page hides the real message; there are two ways to see it — the panel's Reports → Recent log messages, or from your shell:

BASH

drush @example.com watchdog-show --count=20     # Drupal 6/7
vdrush @example.com watchdog:show --count=20    # Drupal 8+

(vdrush needs Unlock Local Drush first and runs from the platform's app root.) Or open the site's .dev. URL to see the PHP error right on the page.

→ Something looks broken

Locked out of a site

The panel's Reset password task mints a one-time login link for that site's admin — no email, no database. Or straight from the shell:

BASH

drush uli      # prints a one-time login URL to your terminal

→ Passwords & access

Blocked, or slow from one network

Can't reach anything from one place? First confirm it's you, not the site: load it from a phone on mobile data, or another network. If it loads there, a routine web-abuse block has probably caught your address — stop retrying (that just resets the clock), switch networks, and it lifts itself in about 15 minutes. Still stuck after that? Send a support request with your IP (search "what is my IP"), roughly when it started, and what you were doing.

→ Blocked or slow

The tasks (and one toggle) most people touch

Task or file	What it does	When to reach for it
Verify	Re-writes the site's config to match disk — the safe "fix it" button	Almost anything odd; safe any time
Flush all caches	Empties every cache so fresh content shows	A change won't appear after ~60s
Rebuild registry	Heavier: rebuilds the Drupal 8+ service container (or 6/7 registry), then flushes	Flush all caches wasn't enough
`redis_cache_disable`	Object cache off for one run, set in the site INI	A Clone or Migrate keeps failing on a reused domain

These four are what most people touch. The full escalation ladder, every panel task, and an 11-row "what's yours to fix / what's your host's" recovery table are spread across the five pages of this chapter.

If something's weird

A task stuck on "Processing" for a long time isn't the same as a failed one — the queue is wedged on the server, which is your host's job. Don't pile on more tasks; open a support request naming the site and task. → Task recovery
A page that hangs, then 502s at almost exactly three minutes is the 180-second request ceiling — a page waiting on a slow feed, API or payment service — not the busy-server protection. Fix the slow piece, or move the long job to the shell. → Blocked or slow
Two different logins, two different routes: your Drupal site's own /user/password link works and is the friendly self-service reset; the control-panel login can't email a reset, so that one is a host support request. → Passwords & access

Going deeper

The whole topic, in depth: When something's wrong — the calm first-stop overview, plus playbooks for white screens, failed and stuck tasks, the three logins, blocked-or-slow, and the two site-health tasks.
Related sheets and guides: Caching cheat sheet for the clear-cache deep dive · Drush basics · Shell & SFTP.
Root-side recovery (for operators): Troubleshooting.