You cannot select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
alex 5d850dc322
git: migrate to repo
migrating from self-hosted gitea to github due to maintenance burden.
2 weeks ago
base base: disable 'abort on errors' in tor setup script 1 year ago
btc btc: upgrade bitcoind from 25.1 to 26.1 2 months ago
files/etc git: migrate to repo 2 weeks ago
keys keys: add my other, newer gpg key 1 year ago
lnd lnd: patch update from 0.17.3 to 0.17.4 2 months ago
ndg git: migrate to repo 2 weeks ago
LICENSE initial import 2 years ago git: migrate to repo 2 weeks ago apply: report a bit more info during an update run 1 year ago git: migrate to repo 2 weeks ago

system updates

the plan is for this repo to contain all system updates, incremental in a form of text/source code. a node periodically runs the script which pulls the repo to receive updates executes the latter then makes changes and updates the operating system.

at the moment, all updates are executed in form of shell scripts. these are error-prone and hard to reason about in a comprehesive way once the codebase gets sufficiently large. the short term goal is to migrate shell scripts to something more managaeble like saltstack but with less resource requirements, suitable for embedded devices without python dependencies.

typical update examples are: upgrade bitcoind, lnd and other services, system packages, improve configuration of components such as firewall. the run sequence on the node is approximately as follows:

  1. fetch updates with a git fetch.
  2. provide a git diff on the screen and confirm with the user.
  3. pull in the changes with a git pull --verify-signatures.
  4. run

at the moment, an on-screen diff and confirmation aren't implemented yet. nd and ngui is where it'll happen, in the ndg repo.

when configuring a new node, clone this repo and set up a cron job to execute the script once a day. The script requires REPODIR and LOGFILE env variables set.

TODO: add a list of supported platforms; the "native" is void linux.

testing a live change

the procedure to run a modified sysupdate on the device while ssh'ed into the instance.

first, make sure periodic updates are disabled:

chmod -x /etc/cron.hourly/sysupdate

then set required env variables and run the apply script:

cd /ssd/sysupdates

to reactivate periodic sysupdates, flip the x bit:

chmod +x /etc/cron.hourly/sysupdate

note that the periodic sysupdate script will revert the repo to the latest commit of the branch specified in the script or master as the default.