sixnix-deskctl

For whoever is running the desk.

sixnix-deskctl runs a desk’s maintenance tasks. It finds the web container and runs there, so the long docker compose exec -u www-data web php artisan in front of everything is not something you type.

curl -fsSL -o /usr/local/bin/sixnix-deskctl https://registry.sixnix.net/install/sixnix-deskctl
chmod +x /usr/local/bin/sixnix-deskctl

Run it from the directory holding docker-compose.yml. If you would rather run it from anywhere, set DESK_DIR to that directory.

It is not sixnix-desk

There are two CLIs and they do unrelated things.

• sixnix-desk works one project’s issues over the API, from anywhere, with an automation key. For CI jobs, scripts and assistants. See Automation keys and the CLI.
• sixnix-deskctl administers an installation, on the machine it is installed on. This page.

What it does

This is sixnix-deskctl help, which is where it comes from. If the two ever disagree, believe the command.

  Everyday
    status                  which services are running, and whether they are well
    version                 the release this desk is on
    logs [service] [-f]     what the containers are saying
    tail                    the desk's own log, followed
    shell                   a shell where the application lives
    tinker                  a REPL with the desk's database behind it

  Backups
    backup                  take one now: the database and the files
    backups                 what backups there are
    restore [NAME]          put one back, replacing what is here. Asks first.

  People
    admin EMAIL             create an administrator, or reset one's password

  Workers
    workers                 the queue workers, and whether each is running
    workers restart         restart them without restarting the container

  Upgrading
    upgrade                 back up, fetch the new compose file and image, and bring it up
    upgrade --dry-run       say what would change, and change nothing

  Upkeep
    migrate                 apply any pending database changes
    migrate:status          what is applied and what is pending
    optimize                rebuild the desk's caches after a configuration change
    cache:clear             empty the application cache
    queue:restart           tell the workers to pick up new code
    queue:failed            jobs that gave up
    queue:retry ID|all      put them back
    mail:poll               fetch from the mailboxes now, rather than waiting
    knowledge:refresh       re-read knowledge sources that are due

  Demonstration data
    demo [--scale=N]        fill this desk with invented brands, tickets and issues to look at
    demo --undo             take it all out again

  Moving in
    blesta:import-tickets   import tickets from a Blesta Support Manager. Try --dry-run first.

  Anything else
    artisan COMMAND …       run an artisan command directly. Everything is reachable this way;
                            the list above is what you are likely to want.

  DESK_DIR=/path            where the desk is, if you are not standing in it

Why the list is short

Laravel ships around a hundred and fifty commands. Most mean nothing to somebody running a support desk, and a few of them end it. The short list is what you are likely to want; everything else is still there behind sixnix-deskctl artisan ….

Four are refused by name rather than hidden, so that somebody reaching for one is told why:

migrate:fresh, migrate:reset, migrate:refresh, db:wipe	every ticket, client and setting, with no undo
queue:flush, queue:clear	queued work, mail waiting to go out, drafts, indexing

Each refusal names the command that would have been the right one instead, and tells you how to insist if you really mean it.

What it deliberately will not tell you

sixnix-deskctl status says which services are running. It does not say whether the desk is well, and the difference matters: a queue with no worker looks perfectly healthy from outside. The desk’s own answer to that is Settings → Health, which watches the heartbeat the scheduler writes.

The reason it is not here: the only command that touches the heartbeat records one. Calling it from status would make a desk whose scheduler had stopped insist it was fine, which is worse than the question going unanswered.