Not sure if it’s just my setup or if it’s a combination of bindings or even interaction between OpenHABian and other installed packages, but I’m finding that my OH2.5 system is unstable. It will suddenly stop doing things – sending notifications or running rules and then when I try to access the UI, it just fails with the browser reporting no such page/page timeout. If I’m lucky, ssh will still work and I can get in to either restart OH2 or reboot the entire box.
Version 2.4, went for months without hiccup. When it did go wrong, it was probably me trying some new rule or binding that I’d got wrong. These days, just leaving it alone causes issues. The most frustrating part is that I can’t figure out why: there’s no consistent method of failure that I can find.
Sometimes I’ll use the openhabian-config to do an apt update and stuff just won’t restart afterwards. The forums tell you to stop – clean-cache – restart and that you might have to restart several times before it will work. My personal record for that “restart a few times” is eight before the ssh connection failed to allow me to re-connect and all access was basically blocked.
Other times, things just stop working – the light’s don’t go on at the right time, or some do and some don’t, even if they are in the same rule implementation. I’ve had the first and third light turn on and the second light not in the same rule block (see below).
//Turn on Sunflowers - Zigbee switch
Innr_Plug_01.sendCommand(ON)
//Turn on Blooms - Hue
Hue_FF_Toggle.sendCommand(ON)
//Turn on Austin - Z-Wave Socket
Light_FF_TableLamp_LoadLevelStatus.sendCommand(50)
So the 2 wall sockets – running against different technologies and using different bindings, work. The Hue doesn’t. Or might. Depends…..
I have a switch (bCancelSunrise) which is set to OFF on system start (or is NULL or is UNDEF), that randomly turns ON with no input from anywhere else. This switch controls the kids morning alarms, so if it goes to ON (i.e. cancel the sunrise), their lights don’t brighten and they don’t get up…… The first I know about it is when I get a notification saying “Sunrise cancelled successfully”.
The logs often show an increasing number of java errors, interspaced by normal operations, but again, nothing consistent or in a recognisable pattern. Prior to 2.5 I didn’t even know that it was even possible to clean-cache (was it?), now it seems to be the default response to any question – stop, clean-cache, start, wait and restart again if it doesn’t work.
As it stands now, I’ve rebuilt the system from scratch roughly every 2 weeks. I’ve now got it stripped down to the bare minimum and can do a new SD card to running system in about 3 hours.
And that is NOT acceptable.
I’ve changed SD cards, booted from new USB sticks and even used different Rpis (all 3B+).
I’ve trimmed down the bindings I use, to try and reduce potential clashes there could be – after all, no-one will have the same combination of bindings, services and devices that I have.
My original v2.4 bindings list:
- v2 bindings – astro, chromecast, evohome, gpstracker, harmonyhub, hue, mihome, rfxcom, squeezebox, tradfri, weather, zwave
- v1 bindings – expire1, mios1, mqtt1, plex1
My updated v2.5 bindings list:
- v2 bindings – amazonechocontrol, astro, chromecast, darksky, evohome, gpstracker, harmonyhub, hue, mihome, mqtt, rfxcom, squeezebox, tradfri, zwave
- v1 bindings – expire1, mios1, plex1
My current v2.5 bindings list:
- v2 bindings – astro, chromecast, evohome, gpstracker, harmonyhub, hue, mihome, mqtt, squeezebox, tradfri
- v1 bindings – expire1, mios1
The significant differences between v2.4 and v2.5
- New MQTT binding has been used replacing the older mqtt v1 version
- Removed the mail action (no longer send emails as alerts)
- New 2.5 versions of my bindings
- Added Jython and new RuleEngine
The biggest change between 2.4 and 2.5 was the removal of the weather binding – replaced by DarkSky and the inclusion of the AmazonEchoControl. I also setup and tested the Jython language for rules. All of these plus the RESTDocs and MyOpenHABCloud have all now been removed without any obvious improvement in stabiity.
The frustration is building – people in the house depend on the system, and if it doesn’t work there will be trouble.
twistednix's 