I’ve been using HA for some time on my rPi3 and its been working pretty well. Recently (after the .115 upgrade I think?) I started seeing a strange bug that is causing me to restore my system every few days.
HA works great, all automations run as they should, until I log in to the front end. As soon as I do that I get the “refresh” button, or “web page cannot be reached” and all my automations stop working.
I’ve tried new SD cards, tested the SD cards, Fully formatted the SD cards. Restore with Wipe, Restore without Wipe, all gets the same error after a few days.
Where do I go from here on my troubleshooting? I’m uncertain how to access the logs if I can’t get into the front end. Thanks for any assistance!
same issue here. Crashes every few days. Also installed on a new SD and one week later it crashed again.
I also have a RPI3 B+.
Noticed Green and Red light kept burning when it crashed this morning.
When restarted (unplugged power) I also get the unsecure webpage stuff and I think I have this after every crash.
Apart from a good quality SD card, the other big player for this type of behaviour is the power source. Are you using the official RPi power supply, or something that is able to supply at least the same?
I bought a new SD card last week : SanDisk 32GB micro SD Extreme Pro A1 U3.
I use the powersupply for years and didn’t have any issues until a few weeks ago that my hassio is crashing 1 or 2 times a week. Config didn’t change a lot so doubt that it is the powersupply. Since Guyanthalas looks to have the same issue very recently it smells like …
I’ve got a few SD cards that I rotate through when something goes wrong. I use an SD Card formatter and an SD card tester if I’m suspicious. So far those have come up clean on two cards and the crashes persist.
I installed the rPi power supply check add-on a while back and verified that the power was clean (after I removed the super sweet power button thing in the middle of the power cable. Lame. )
At this point I’ve gone back to the very basic config only thing. I’ve removed all my addons and many of my integrations (Like HACS).
What I’m most disappointed with is that there is not an easy way to help pinpoint the issue. I can’t recover a log file after a crash, and I am not certain that a log file would do me any good anyway unless there was some flagrant error. Hell, even a “debug mode” would be useful in a case like this I would think.
The reason I mention a power supply is that recent releases of HA have made changes in templating that have affected some users and caused an increase in computing demands. Therefore, a power supply that works but is at it’s limit, may not be able to cope. That’s only very specific templates that have had this issue though and some improvements are likely to come in HA 0.117.
Do you have the SSH addon installed? If you can install that before the system crashes, you can use a program called WinSCP to see the log file, but only if the Host System is still running.
In WinSCP keep the File protocol as SFTP, use your Pi IP address as Host name and use Port number 22. The default User name is root. When you have access WinSCP brings up the folder structure of your Pi in a File Explorer like window, so you can drill down to find the log file, which is in the config folder - home-assistant.log. You would probably need to change your logger log-level to debug. It might still be tricky to pinpoint the issue though but at least you’ll have more detailed info to post on here. Seems a sensible idea to strip everything back and reintroduce integrations though.
I hope you find the problem, I understand your frustration since we become very dependant on these systems. I’ve moved to an SSD and don’t use an SD card at all, after quite a few card crashes. I’ve got a RPi 3 but a fairly small number of integrations compared to a lot of users. Fingers-crossed, it’s been running without an issue, although I’m now concerned I’ve just jinxed myself!
For sure, and I’ve got a dedicated 2.5a power supply for the thing.
Yeah, I have SSH installed but once it crashes I can’t seem to use putty. Haven’t tried WinSCP though, so I can give that a go if it continues to fail. I’ll look at changing the logger level after the next reinstall.
Eventually I’m going to move to an SSD. The drives are cheap enough, but my understanding was that there was some sort of complication getting the whole kaboodle to boot from an SSD (and additional hardware needed, at the very least a USB3 enclosure).
Unfortunately if Putty doesn’t work then WinSCP won’t either. Hopefully you can find something in the log but debug mode will write a heck of a lot more to the SD, so be sure to turn it back to warn at some point.
You don’t need USB3 enclosure for the SSD, I did but Pi 3 uses USB2 anyway, so there won’t be any performance improvement. There’s a list of recommended hardware on the forum somewhere. I followed this guide https://www.youtube.com/watch?v=uuGz9d_tTzc&t=1s
It’s a pain to do but sometimes a nice fresh start is the way to go. Best of luck solving your problem.
Today hassio crashed again. So around 10 days now
It crashed while upgrading zigbeet2mgtt addon. Green and red led keep burning.
I’m still on 0.116.4 and was upgrading all the add-on’s with updates before upgrading i wanted to upgrad to the 0.117 version
Right now I’m still in testing for my setup. I’ve wiped the unit clean and have been putting everything back in one at a time manually (no snapshot restoring). Right now I’m running super minimal. If it STILL crashes on me in a few days I’m going to suspect it’s the Pi.
If anyone is following this, I just wanted to update that so far I’ve chosen to go back to a “start from scratch” style installation which appears to be stable. I only have about 25% of the items reinstalled that I had working before, but that is pretty strong evidence that either an integration or custom component was causing my problem (could easily be coupled with an HA version switch). Either way, I think my issue is “solved” to the best it can be.
Still wish there was a better way to retrieve errors after the fact when it comes to a crash, but I’m sure as HA continues to grow up we’ll see that implemented eventually.
