Shells and logs

The SSH user you connect as picks where you land:

1ssh cd.pico.sh              # a shell in your projects directory
2ssh web@cd.pico.sh          # a shell in a project
3ssh web@cd.pico.sh ls       # a command in a project
4ssh c-web-web-1@cd.pico.sh  # a shell in a container
5ssh l-web-web-1@cd.pico.sh  # follow a container's logs
6ssh a-web-web-1@cd.pico.sh  # attach to a container

Containers are named as compose names them: <project>-<service>-<n>. status <project> lists them.

You always log in as your own account, the one your SSH key belongs to. ~<owner> in the SSH user picks whose projects you act on, such as a team's, and whoami shows both:

1ssh '~pico@cd.pico.sh' whoami

Shells and commands in your environment get the same as variables: PICD_USER is your account, PICD_OWNER whose projects you're in, and PICD_PROJECT the project, in a project's directory. Shells and commands in a project also get its environment variables.

Shells that resume #

Your shells run in tmux, so a dropped connection doesn't end them. Your next shell in the same place, your projects directory or the same project, attaches the shell you were in, with whatever it was running. A second shell while the first is connected gets a new one.

tmux's status bar shows the project you're in and your windows. tmux's keys start with Ctrl-b:

The mouse selects panes and windows, and scrolls. ~/.tmux.conf changes any of this. A shell left detached for a week is closed.

To use plain shells instead:

1mkdir -p ~/.config/picd && touch ~/.config/picd/no-tmux

Updates to your environment end its shells. picd holds an update while you have pushes, deploys, or commands running, and for up to 6 hours while only shells and logs are open. Commands run in a terminal, with ssh -t, count as shells. Your shells get a minute's warning first.

Your home directory #

Your home, ~, is kept in /data/home, so your dotfiles, ~/.ssh keys and config, ~/.gitconfig, and docker login credentials survive restarts and updates. It starts with picd's default shell setup, which you can change like any other file. Webhook deploys use the SSH keys and config there.

Updates to picd's default setup don't change a home that exists. To start over from the latest default:

1ssh cd.pico.sh home reset

Your previous home moves to /data/home.old, replacing an older one there. Copy back what you want to keep, such as ~/.ssh.

Logs #

logs shows a project's last 100 lines from every service, or from the services you name, and -f keeps following them:

1ssh cd.pico.sh logs web
2ssh cd.pico.sh logs web db -f

Your docker engine #

Your shell has docker and docker compose for your own engine, so you can inspect, restart, and debug your containers:

1ssh cd.pico.sh
2docker ps
3docker logs --tail 50 web-web-1

compose <project> <args> runs docker compose in a project without a shell, with the same compose files a deploy uses:

1ssh cd.pico.sh compose web ps -a
2ssh cd.pico.sh compose web restart db

Your shell runs as root beside your engine: it sees every process in your environment, as htop shows, and shares its network. That also lets it stop docker or break its networking. If your environment gets stuck, restart it:

1ssh cd.pico.sh restart

Everything in it stops and starts again, with your repositories, projects, volumes, and home kept. Your shells end, and your apps restart.

To stop your environment and its apps until you next need them:

1ssh cd.pico.sh shutdown --confirm

Without --confirm, shutdown only says what it would do. Your data is kept, and the environment starts again the next time you connect, push, or run a command that needs it. Add --delete to also remove the environment itself, so it's created afresh when it next starts.