I run HA on a Raspberry Pi 4. When I do a core or OS update, it can take 20 or 30 minutes. As far as I can tell, HA is functional for some of this time, but not all. (It may depend on which sort of upgrade is happening, and may mostly have to do with the REST api, since I have external scripts that notice the outage.)
First, is this normal? Expected?
Right now I’m upgrading directly from Core 2025.9.3 to 2025.10.3. It’s taking well over half an hour by now. This may be the first time I’ve done such a big jump – does the size of the version jump affect the duration of the upgrade? (Do the intermediate versions all get installed behind the scenes?)
If I were running HA on a different platform, like an x86 board with a “real” disk, would the upgrades be much faster?
When i ran HA on RPI i didnt experience this, i used sd card. I dont thinks its the fact its an sd card more so a bad one with low write speed. regardless ditching pi all together for an intel nuc would benefit you @pgf
Interesting – I guess I should take a look and see what SD I’m using. I thought it was a top of the line Sandisk, but it’s always possible I was careless and used a different card on installation day. I can also do some simple speed testing.
Damn. Did some quick tests using dd. Looks like I screwed up and used a slow card: write speed is 6.0MB/s.
Edit: But when I went to look at the physical card, I find that it is, in fact, the Sandisk Ultra 32GB Class 10 card that I thought it was. Which should be capable of much faster speeds than that. In fact, I have another RPi4 running linux for my CNC machine, using the same card, and it gets 22MB/s using the same dd write test. So just getting a new card for the HA machine probably isn’t the answer.
Question: I did the dd test from the login shell I get when I ssh to the HA box. Is that a good test, or is that in some virtual container several steps removed from the real hardware?
No there shouldn’t be to much latency with the docker layer.
Can you check if the SD is still healthy? It might just be that the speed is slower since it is encountering (and trying to fix) errors.
Well, much faster not really, class 10 defined 10mb/s read & write. It might be just that less and less writable flash cells are available on your old card and typically the write speed depleds torward the end of life. The end usually is a read only card (grid lock).
If you use a card for application or OS watch out for the correct A2 rated card (like the docs tell us!)
No. But one update can significantly take longer than others if a db update/migration is needed
I have found HA updates taking a loooong time (about 20 minutes) too, on my x86 system.
I have worked out that (in my case anyway) the backup before the HA upgrade is taking almost all the time. Just like Windows, each update adds functionality (most of which I don’t use), and takes up more disk space.
I found a small form factor PC on ebay with an mSATA SSD that I hope will help. It’s a better form factor for my equipment rack as well, which is nice. And it will free up the RPi4 for other tinkering projects.
Sigh. Got the PC. It’s nice – 2Ghz Celeron J1900, dual enet (important to me) + wifi, a couple of serial ports, 8Gb ram, 240G ssd. Unfortunately, I can’t get it installed. I get the “Slot A (OK=1 TRY=1)” grub screen, and while it does boot past that, it gets to the ha cli page and that’s it. Glad I wasn’t desperate for this transition. :-/ UEFI is enabled, secure boot is disabled, I’ve tried recreating the install media, none of those helped.
Edit: I totally forgot that HA admin is done from the web, and the cli screen didn’t really make that clear. It’s working just fine!
Just to cap off this thread: I’m in the process of transitioning to the new PC, and as it happened, last night both the Rpi and the PC wanted updates from core 2025.10.3 to 2025.10.4, so I did them both, and timed them, roughly. The new PC took about 3 minutes. The RPi, in service for going on 3 years now, took 35 minutes. No pre-upgrade backup was done in either case. So clearly my SD card is shot. All of the small writes (database and logs) that HA does are likely the culprit. I’m glad I’m retiring the RPi for other reasons as well, but if I were keeping it, I’d definitely be thinking about other storage solutions.
Thanks for updating!
In the last week alone I think I came last at least 3 threads where the SD card came out as the culprit. 13 years in with Raspberry PI’s and the same issues persistent. Consistency is key…