Reactor-Editor / Reactor-ui-Status Errors

[toggledbits]

@Cadwizzard said in Location of the rules on filesystem (.json etc) in HA Add-On:

So something isn't saving the live configuration properly somewhere.

Any time you back up a running system, you run the risk that data hasn't been flushed from cache to non-volatile storage. This applies to everything, not just Reactor.

[Cadwizzard]

@toggledbits That I understand.
But in case I was unclear above. I had a lot of saved and working global expressions that were lost after a host restart. They referred back to one test expression I'd previously created. I then went to the backup to find that was the same

Is there a way I could update this docker container myself to the latest version of msr. The add-on maintainer seems absent and I don't want to fall behind.
Docker is not at all my speciality but I can follow guides/instructions.
I used docker-compose on the pi, but on the HA Yellow it looks like HA OS is is protected and wipes out anything not in the homeassistant directory or the installed plugin containers (including when I installed the docker-compose plugin).
Would it just be a case of logging into the container and running the normal docker pull commands, or running them outside the container but specifying the docker container in the command?

Many thanks!

[Pabla]

Don't mean to highjack this thread, but when I do a reactor backup, reactor is usually running when the backup occurs. Should I instead shutdown reactor then back it up?

[toggledbits]

@Cadwizzard said in Location of the rules on filesystem (.json etc) in HA Add-On:

Is there a way I could update this docker container myself to the latest version of msr.

I wouldn't know (I didn't create the HA Add-On and I don't provide support for it), but maybe @mrw298 can produce an updated version?

[Cadwizzard]

@Pabla I've been running Reactor for years across many platforms and never ever had an issue with backups doing weird things - until running the add-on install.
I never saw it with a docker install as per toggledbits installation guides.

This issue may be just related to the add-on as i've never heard of expressions disappearing before, until here and my install: https://github.com/mrw298/hassio-reactor-addon/issues

[Cadwizzard]

@toggledbits @mrw298 was last showing online here in March.
Thats why I was reaching out in this forum for wider experience

Do you or @Pabla happen to be able to answer a more generic question about Docker in the context of HA (and i'll run the risk myself of something going wrong).

Kind of forgetting that MSA was installed via the add-on, it currently exists in a docker container in HA yellow, which i can see and access via SSH
Should upgrading it mean running commands inside the container, or somehow specifying the addon container by name to update MSR inside it?

I'm really trying to learn more about how docker updates/linux inter-relate, but MSR is literally the only use case i have for it. I'm not a linux specialist, more of Windows sysadmin with vmware etc and only touch linux/docker for Home Assistant/MSR and PiHole, so there is a lot of trial an error with this for me It's not yet clear about the possibilities or the best way to interact with docker after someone/thing else has created containers like in the add-on.

@toggledbits btw your guides are actually great when installing MSR from scratch, thanks for taking the time with the documentation, it really helped previously move from Vera to Pi and keep MSR on the same hardware.

Any tips really appreciated.

[toggledbits]

You could attach to the running Reactor container (typically) using docker exec -it reactor /bin/sh. This assumes that the container name is reactor; you may need to use docker ps to establish the actual container name, or alternately, use its ID in the docker exec command.

From there, you should be able to go to the directory above the container's running Reactor: cd /opt

Then, you can download the latest Reactor:

    wget 'https://reactor.toggledbits.com/download/latest/<image-tarball-name>'

Then, un-tar the image in place: tar xvzf <image-tarball-name>. This will overwrite the Reactor install files native to the container with the files from the downloaded tarball.

Restart the container and it should be running on the installed/updated version.

A word of caution: This process only modifies the running docker container. It does not modify the image. I have no idea how HA handles the containers for add-ons. If the container is destroyed and recreated on every startup of HA or its add-ons, then your changes will go away, because the container is being rebuilt from the image, and the image has the original as-built Reactor files on it.

[Cadwizzard]

So, @toggledbits - thanks again.
It IS possible to update the docker container. It does survive a Home assistant restart..... but not a host reboot. The version reverts. So the container is rebuilt.

I've built a test Pi with HAOS on it for testing outside of Prod.
I created one test rule and one global expression while testing the updates.

Host reboot wiped out the global expressions again. @mrw298 There is definitely some fault with the add-on, this is a completely seperate install and hardware with no config, except the SSL and Reactor add-on installed.

I guess this leaves me to try to figure out where is safe to manually install reactor in HAOS and get round the "Error response from daemon: error while creating mount source path '[any path]': mkdir [any path]: read-only file system"
Update-
Actually the whole point of moving to HA Yellow was to be on a fully supported HAOS - So I'm going have to learn how to build my own local add-on to pull reactor into a supported container. My need is to run reactor on the same hardware as HA, and not have to rely on other plugin maintainers to stay in a supported MSR version

@toggledbits - have you considered building a Home assistant Add-on for Reactor at all?

Thanks again

[toggledbits]

@Cadwizzard said in Location of the rules on filesystem (.json etc) in HA Add-On:

have you considered building a Home assistant Add-on for Reactor at all?

No. There's more to it than building, as you are finding out. That also means I have to support it, including understanding how to do remote troubleshooting and debugging in that environment. It's not on my radar right now.

[Cadwizzard]

@toggledbits thanks.
I've modified the build.sh from the other add-on and set up docker hub. Despite not fully being able to understand what and where Linux read/write errors actually arise from, I've got the files built and pushed. Had to use a pi to build the arm and aarch, then Ubuntu in Windows to build the amd64 due to errors and warnings about architecture. Unsure if that has to be that way, or they should be buildable from one platform or not.
This was the better idea as when I tried to update the other add-on container, msr warned about node.js being out of date by doing it that way.
I've got my version loading from a local plug in and pulled/running from GitHub now on my test install. Only thing to work out
Only thing left to work out is why the alerts/entities aren't updating, but I'm thinking that's because the default in msr points to a dns name for HA. Which is likely resolving to my prod HA Yellow (incorrect login alerts on there also point to that). Was a rush before I needed to head away. I'll backup/remove and pull the new add-on to my prod. hardware once at a computer I can work properly from, then copy the rulesets etc into a backup for that and re-export it.
Appreciate the tips from you and @Pabla
It's a bit more work to build the files and pull into my own add-on but it's better I have the control and still run a supported HA install

Update: That was it, installed/updated Reactor on my HA Yellow, entities updating as expected