homeassistant

I try to set aside time to do maintenance once or twice a week. I do a backup of my system using the standard backup tool then do the full upgrade of everything, starting with all the HACS and Addons, then the core OS stuff, then reboot. Then I test a few things and make sure it looks OK. I haven't had any issues yet but I know it is a matter of time. I have a second computer I will be testing restoring onto shortly, so I will make that part of my process. No point having a backup if you haven't tested it.

My systems do automatic update checks, but require manual approval to install them.

Though I generally install them pretty soon after seeing that the update is available, so there's not much time between the update being discovered and being installed.

For applications it's pretty much immediate but for bigger things like operating systems it can be a little bit

Update policy?

I see update, I click update!

ROLL DICE BABY!

it does backups... somewhere... I think...

Starting in 2023 (and updated over the years) I wrote an automation to auto apply updates every few days based on a bunch of criteria:

• consider all updates: HA, HAOS, add ons, HACS updates (I’d included HACS updates myself before they had integrated with the update framework)
• update is more than 1 or 3 days old (I’m not interested in being an early adopter for the most part)
• ’sensitive’ vs. standard updates (I wait longer for updates to HA, HAOS, Zigbee2MQTT, ZWave2MQTT, few others)
• order of updates, like apply standard HACS updates first, then update others
• logs everything it does
• keeps state, so if there’s a version bump between installs it delays another slot before considering to apply it
• skip all esphome updates by default now (didn’t initially, but don’t see the value in pushing updates to the devices unless there’s a material gain)
• I have controls on it so like if I’m going out of town I pause it until I’m back just in case

I usually get core HA updates to the y.m.2 version around mid-month so I skip teething pains of most new releases. I’ve had to do a roll back once - I’m still not sure why as I just reapplied the update manually and it worked. It has meant that I’ve needed to update parts of my configuration as HA has progressed, but for the most part that hasn’t been a lot of upkeep.

I think HA is still such a moving target, and especially now with AI beating on everything security wise it’s even more important if you expose it to the world. Given HA’s role leaving it unpatched could be opening your home up to being abused.

HA is still a bit of a hobby project and needs a bunch of care and feeding. I do hope that there’s a point when the dev team doesn’t need to reachitect and refactor configuration. Eventually I hope it’ll be more like phones where updates are relatively transparent.

Didn't know there was an auto update?! I run them on a random schedule when I'm bored. While commuting with the bus and no one is home. Or while watching YouTube videos in the evening.

I use Watchtower to automatically fetch the new Home Assistant container image, (with a 3 day delay to decrease the chance of supply chain attacks) then stop the running container and replace it with the new one. This has been running for about a year and I’ve only had one or two breaking changes. Watchtower is scheduled to run at times when I am usually available to correct anything if needed.

Version are pinned in code. Helm chart, image, RKE2 and OS. Let fleet/argocd/helm handle most upgrades. Infrastructure/management upgrade (Rancher app or rancher cluster components, Forjeo, Harvester, keycloak, etc) get a dedicated virtual testing cluster spun up and manual and automated tested (I don't have good virtual networking testing yet once I do KubeOVN only that might be truly possible). Renovate runs to automatically update all the version numbers.

Checks every 15 min for apps, every hour for infra.

Id love to try to get a web hook from my security scanner to push for an update sooner if a VEX item is identified, but shrug

Mind you, I like running my homelabs, the idea it serves anything useful is just to prove I can lol. They go down all the time too because I'm monkying with the hardware config ALL the time.

I click it when I see it in the app. But knowing there are auto-updates, I might just turn those on instead.

I have regular backups of the entire system, so if I notice something broke, I can always just roll back.

@JelleWho I'm running Home Assistant as container with podman quadlet.
I'm using an Ansible Playbook, which is executed manually for updating of Home Assistant.

If a new HA container image has been fetched, the podman quadlet is stopped, a backup of the container volumes is executed and afterwards the podman quadlet is started again, using the new container image for the container.

For apps and integrations I run then when I would have some time to do a rollback, mostly this is quite quickly. For home assistant updates I always read the changelog and estimate if there could be complications, if so I wait for an x.1 patch to be available before taking the plunge.

The HACS updates I always leave for a few days and read though carefully what has changed, quite often even way deeper into the code.

For Esphome I run a full compile about every year, and I updated then when I change something in the YAML. These run on a Vlan and not exposed, so they are not a big risk I think to be outdated.