I wrote how I solved it, with samba and no restore of HA core.
2 Likes
Restore bricked my home assistant… I am SO BEYOND PISSED OFF at the moment…
1 Like
Well that was fun… Backups that don’t actually restore to a previous state, when downloaded you cannot open them, one that bricked my entire setup…
I’m finally back up and running. And using @Staedtler info managed to import my old flows… Still have complaints about some nodes now not being installed. But at least I am not litterly sitting in the dark anymore (my studies lights are fully managed by HA).
One of the big issues I also ran into, most backups didn’t seem to backup the flows. Esp. not the NodeRed update backups for some reason. I had to dig deep to find a backup file that worked and am glad I’ve not been doing much changes recently.
Lession here is: Do not trust in HA backups…
1 Like
Don’t trust any backups. Test them regularly to make sure they work. The worst time to discover they don’t work is when you need them.
1 Like
The thing is… if backups from HA are so unreliable… What else can you rely on? I backup a lot, almost any change and also store them in the cloud. But discovering that a ‘recovery’ of a full backup doesn’t actually recover the old situation at all worries me a bit. Don’t know what other line of defense there is? Doing a manual backup and recovery before every single update in HA just to check if it works? And how do I know… Discovering that the recovery did ’ restore’, but things remained unchanged… thats hard to check for unless there is an actual change…
I don’t think backups from HA are necessarily unreliable - they may be more finicky to use for restores because of the different environments and add-ons. Therefore, it is important to test the restore process for your environment.
My situation: RPi4 using SD card.
At the end of every month, and before I do any updates/upgrades (I usually wait until mid-month to do upgrades), I shutdown the RPi4 and remove the SD card (let’s call it “production”). I make a backup of the SD card using Win32DiskImager (backup). I insert the backup into the RPi4 and boot. I have just confirmed that I have a good backup AND I store the working “production” SD card in a safe place - I know I have two working copies (I actually have 4 because the saved “production” is actually rotated with 3 other saved “production” copies from previous months.
I can always drop back to the saved “production” SD card if I discover something strange or wrong.
I also use Add-on: Home Assistant Google Drive Backup so I have an off-side file I can use if needed (I only test this once or twice a year since my SD card rotation strategy works well).
If I was using an SSD I would create a disk image (not an HA backup) of that and store it in Google Drive and another cloud provider then I would restore it to see if it works.
This may be a little time consuming but It is better than struggling through trying to figure out why my back can’t be used to do a restore.
I think it is more convenient for me to backup, test, restore using the entire disk image than using the built-in HA backup method.
1 Like
Wow for a novice user this is all very confusing. I came back home to find it non-functioning and when I opened node red it was completely empty. I immediately restore a full backup from 2 days ago and everything is fine now (node red 15.1.1 up and running fine).
My addon_configs is completely empty and I disabled node red auto update. What should I do update to 16.0.2 ?
I actually do use an SSD. This may be a good idea. I also offload my backups using that Google Drive Backup. But am fully reliant on HA’s backups. What scared me is that when I restored a full backup, it wasn’t actually restored to the old situation (i.e. pre 16.x), when downloading them from either Google or HA directly I could not open all files. And iceing on the cake was a restore that just caused HA to never come back up… Eventually found an older backup of 7 days that at least had the flows and restored those.
Edit:
This has me baffled… why I can not open these files. Also still thinking something wrong with my laptop or setup?
1 Like
Not impressed at all, to say the least.
1 Like
Thanks @JBfr! This is as simple as it will get - and it works.
Just be aware that you will revert to the state just before your very last saved edit. So your latest edit has to be re-done.
2 Likes
This is not the lesson.
First understand Why the change.
At the end of the day this change means your flows and such WILL NOW RESIDE IN THE BACKUPS OF NODE RED post 16.0.2 not HA core. So had the bug not happened you’d all be extatic now because ultimately this will eliminate most of the confusion we’re dealing with. It just unfortunately happens at. The worst possible point.
Second.
In a real IT shop if you go to restore and it won’t because the backup is corrupt. Its YOUR FAULT (i have literally fired someone for this VERY thing. Didn’t test the restore process and validate the BACKUP leading to data loss.). In a shop we don’t trust ANYTHING. Not people, not software, not storage. (Hell we call current security methods ‘Zero Trust’) Fortunately there were multiple ways to get to the data. I test my backup about every 4th one (about a month) just by cracking open the tar to verify it extracts. The way HA is built if you have the files you’re more than halfway home. And I never delete the one I tested until another I’ve tested is there to replace it.
So. Full 1 (tested) , Full 2, full 3, full 4, full 5 (test), delete Full 1 (unless it’s my semi-annual or year end.
Technology is never a panacea and sometimes (me included I was WAY off guard for this one this morning) we get bit.
Your lesson is don’t trust TECHNOLOGY without validation.
5 Likes
I strongly suspect the files are corrupt unfortunately. What other backups or things are available?
Home assistant backup files are encrypted in a way that prevents them from being opened by the regular tar executable.
Check out: decrypt-ha-backup · PyPI
A related thread here: Unpack encrypted snapshot files
Woooow what a day ! Server is up again with release 16.0.2 but can’t find my nodes. Do I have to open a backup file, if so how do I do that ?
Win11 opens them natively
1 Like
I’ll give a try, thanks. But looks like my NodeRed backups are wrong, they only do 12kb lol
Just a dummy question…
if I do a full restore with my backup 24h old, it will work right? I will get all my flow back?
cheers
1 Like
If you do a FULL restore that includes your NR config and restart after yes.
Your NR config pre 16.x is currently in your HA core backup and your node red addon is in the NR Addon Backup. So you’d need to make sure BOTH restore. I’d you miss one or the other then you can get into the ‘wheres my stuff’ again loop.
Post 16.0.2 everything you need will be inside the NR addon backup.
1 Like
Nearly there, got my nodes back but looks like my Mosquitto server doesn’t communicate with my NodeRed anymore, damn
1 Like
I got lucky in that the upgrade did not overwrite the .flows.json files or I would have really screwed. The old files were renamed to flows.json.backup or something like that so I was able to recover by importing them back into a new instance of nodered.
My backups didn’t work either. I set my backups to take one full and three incrementals weekly, and overwrite the oldest after the 4th backup, because space is tight. But when I upgraded to 16.0 and it failed, I tried to upgrade to 16.01, then.02 and so on, which backed up the older instance each time. This eventually overwrote the oldest 15.11 backup so I lost it.
@NathanCu being a fellow IT guy myself, I completely agree with you when you say backing up, and testing the backups regularly is important, because we know the consequences of what a failed backup could mean in a professionally managed environment. However, most of the these guys aren’t IT people, so they (including myself) have a certain degree a faith that if something goes wrong, we can simply restore the last known good configuration with a click of a button. As a somewhat poor analogy, when we drive our cars, we don’t do a complete checkup on the engine every time we go for a drive. We have to have some faith that it will get us to where we need to go. And in such a non-professional non-production “hobbyist” environment, that is the way it should be. Things should just work or at least have a way to recover easily if it doesn’t.
HOWEVER, and not to go too hard on Frenck (or whomever created the Nodered addon),publishing a flawed release of a software with such a disastrous consequence as causing a virtually bricked automation system that so many people have come to rely on is absolutely “unprofessional” (for the lack of lighter term).
Luckily there is a fix for this that will probably work for many people, but it could have much…MUCH worse and rendered MANY people without working automation, which many (including myself) have grown to depend on.
I may speak for myself when saying this, but please PLEASE double,triple…quadruple test your applications before releasing to the public. Many are not IT savvy nor do they know enough about JSON files or accessing the addon share to make changes, nor should they. I’d like to think that HA has evolved to the point where its ready for anyone who is interested in this stuff to adopt it, but after today, I am beginning to have my doubts, and all releases should come with a forewarning of potential damage.
3 Likes