KellerFuchs

Rationale

Older versions of hashbangctl(1) used to be called through sudo(8), with all users being allowed to call that command without password.

This is not optimal, both for security reasons and because it would make it very complex for hashbangctl(1) to outlast the user's session (for instance, in the case of user deletion).

Alternatives

KellerFuchs

_hostname=( bigsqueakerbox compooter lightningtrappedinpoisonedglass )
_user_1=( loud soft big smol kyuuuut )
_user_2=( shekken shekkie squeak )

sample() {
    local n=$((RANDOM % $# +1))
    echo "${!n}"
}

prompt() {

KellerFuchs

Keybase proof

I hereby claim:

• I am kellerfuchs on github.
• I am kellerfuchs (https://keybase.io/kellerfuchs) on keybase.
• I have a public key whose fingerprint is 700E 88FC 14B5 3311 1AD7 A40F 1D7C 08C6 CC26 3D74

To claim this, I am signing this object:

KellerFuchs

#!/usr/bin/env bash

function silent_ssh {
    local connectionString="$1"
    local commands="$2"
    if [ -z "$commands" ]; then
        commands=`cat`
    fi
    ssh -T $connectionString "$commands"
}

KellerFuchs

# -*- eval: (ansible) -*-
- apt_repository:
    filename: "{{ item }}"
    repo: "{{ packages[item].repo | mandatory }}"
    update_cache: no
  notify: apt update

- apt_key:
    keyring: "{{ item }}.gpg"
    id: "{{ packages[item].key.id | mandatory }}"

KellerFuchs

#!/bin/sh -e
# Adapted from the Hashbang userdb testing script:
#  https://github.com/hashbang/userdb/blob/master/test.sh


run() {
    normal='\e[0m'
    yellow='\e[33m'
    printf "${yellow}%s${normal}\n" "$*" >&2

KellerFuchs

[Socket]
ListenStream=127.0.0.1:6666
Accept=true

[Install]
WantedBy=sockets.target

KellerFuchs

/.vagrant
/log
*.retry

KellerFuchs

# Let's say you only want to connect to tilde.town through #!,
# for example because the firewall at work blocks tilde.town
Host tilde.town
  User lycaops
  ProxyJump sf1.hashbang.sh


# Your #! configuration
Host *.hashbang.sh
  User opslyca

KellerFuchs

Inspired by @alicegoldfuss: https://twitter.com/alicegoldfuss/status/997268087429713922

Her palms are sweaty, knees weak, arms are heavy
There's errors on her console already, kubernetes
but on the surface she looks calm and ready
to redeploy, but etcd keeps on glitchin'
losing her commands, the whole node grows so loud.
She opens a shell, but the pods won't go down.
The SLA runs out, time's up, outage, blow!
Escalate to high severity, oh there goes availability.