GW1 with FW 25.10.4 Disconnects. Perhaps Add Notifications?

[aesculus]

Since my GW1 updated its FW to 25.10.4 I have been having occasional disconnects with the GW. Even the Tesla app gets no current data and the system is in Standby state.

Going to the Tesla app WiFi setting I see that none are selected. I have a WiFi extender that uses its ethernet output to connect to the GW and the GW has also been set to use WiFi secondarily. Of course, now that it is at 25.10.4 I can no longer see these settings but its been that was for a long time.

If I go into the Tesla app and select the WiFi setting as the WiFi it is set as a backup, then it goes out of Standby and all is good. After a few minutes it moves back to the Ethernet IP address.

For the life of me I cannot figure out why it behaves this way. I would contact Tesla but I think I would be wasting my time.

The first time it did this it was 36 hours before I knew it happened (Netzero gave me a notification of no or low solar), and I lost all of that periods data. It does not show up in the Tesla App either.

The second time it happened I discovered it by accident and it was only 3 hours and it appears the GW just flatlined that segment.

Is my situation unique and perhaps I have a HW issue or just a issue with GW1 and FW 25.10.4?

I was wondering if the PWD could be enhanced to provide some sort of monitoring and alerting for a long disconnect so we can respond to these types of events?

[jasonacox]

Hi @aesculus this is a known issue with Firmware 25.10.x - Please see #637 for details and possible resolution.

For your setup, I would recommend removing the bridge you created. It is possible that the Powerwall is trying to use the ethernet connection to get to the internet, which I suspect your WiFi extender doesn't allow. This could be the reason that the Powerwall loses connection.

I was wondering if the PWD could be enhanced to provide some sort of monitoring and alerting for a long disconnect so we can respond to these types of events?

Interesting monitor alert idea. This tool by @mcbirse might help: https://github.com/jasonacox/Powerwall-Dashboard/tree/main/tools#powerwall-status-monitor-and-api-request-dumper or at least be a starting point.

[aesculus]

I think my problem is either different, or because I have a GW1, my symptoms are. Here is what is happening best I can tell.

The GW is dropping both the WiFi and Ethernet connections randomly. I can restore the WiFi by using the Tesla app and going back in and selecting one of the available WiFi connections. It then starts reporting to both PWD and to the Tesla app.

After a few minutes it will automatically switch back over to the Ethernet connection I have through the extender. Note I don't have any fancy RPi hookup because I don't have TEDApi because on am on GW1. And looking at Teslas current connectivity documentation, Ethernet should still be a valid way the GW communicates to Tesla via the Internet.

I would love to work on a alerting solution, so I will look at the link and hopefully be able to care out some time to dive into it.

[jasonacox]

Thanks @aesculus ! I forget about the GW1 behaviors. Please feel free to contribute back any alerting solution you may find or develop! :)

[youzer-name]

For what it's worth, my recommendation for alerting is to use Uptime Kuma which can be run under Docker and is extremely versatile in the monitors and notification services that can be set up. Since PWD runs under Docker, anyone using PWD is already set up to install Uptime Kuma. You can set up notifications by email (SMTP), Discord, Slack, Microsoft Teams, Pushover, Pushbullet, Telegram, Twilio, custom webhooks and dozens of other methods.

I have a ping sensor set up in Uptime Kuma that pings the gateway and 192.168.91.1 and fires two separate notifications when the ping fails. The first one goes to my phone via my self-hosted Gotify system. The second alert is a custom webhook that triggers a flow in Node Red. That flow connects via SSH to the travel router I'm using to bridge to the TEG network, disables the WiFi interface, and re-enables it 10 seconds later. This automatically resets the connection when it drops. That could all be done directly in Node Red, since I could ping the gateway from there, but I've tried to consolidate as much of my monitoring as possible in Uptime Kuma.

The instructions for installing Uptime Kuma are here: https://github.com/louislam/uptime-kuma/wiki/%F0%9F%94%A7-How-to-Install

I currently also have alerts set up directly in Grafana using its built-in alerting system. It checks the database and alerts me if there is no recent data, so it will alert me to a data issue even if the pings are good.

As I was writing this I realized that I could use Uptime Kuma and the InfluxDB http API instead. I find everything about the alerting in Grafana to be over-complicated and difficult to set up, so I'm going to disable that and switch over to Uptime Kuma right after I finish this post. One of the monitor types in Uptime Kuma is a http request with a check for a matching keyword in the response. I'm going to use:

http://{influxdbIP}:8086/query?pretty=true&db=powerwall&q=SELECT%20%22home%22%20FROM%20%22http%22%20where%20time%20%3E%3D%20now%28%29%20-%205m 

as the URL for the http request. If the response contains the keyword "values" then the data is good. If the response does not contain "values" then it will alert me.

Once you have Uptime Kuma running you can go a bit crazy with the monitoring and alerting if you want. I have Uptime Kuma running on two different hosts (so each can monitor that the other is up) and about 30 separate monitors running.

So rather than trying to add alerting to the project, my $.02 is anyone who wants to monitor the status of the PWD should run an Uptime Kuma instance.

What if everyhing is down? That's where healthchecks.io comes in. It offers free cloud-based alerting for system outages. It listens for a call to a specific https URL. If that request doesn't arrive for a configured amount of time, healthchecks.io can notify you via various methods. And yes, you can set up a 'monitor' in Uptime Kuma that will call the healthchecks.io URL, so if Uptime Kuma (or its host or your internet) goes down, the healthchecks.io alert will fire.

[aesculus]

Just Wow. Will look into it and please continue to post your results as you add and adapt the system.

[aesculus]

OK. Hate to be stupid but I have Uptime Kuma running in a container and I have created an ID and logged in.

I tried your monitoring URL on a browser wand it returns :"values" json.

I added the url into UK as a new Monitor and just took the defaults.

UK complains about the URL and says it times out. Can you give me some hints on how you set the monitoring URL in UK?

[youzer-name]

The URL is basically entered just like I wrote above. I have DNS entries set up on my network, and have "influxdb.lan" as a host name that points to the IP of my PWD/InfluxDB host. My URL is set like this:

Do you have the method set to "Get" on the monitor setup?

[aesculus]

Yes to all. This is what I have for the server link modifying what you posted:
http://192.168.1.108:8086/query?pretty=true&db=powerwall&q=SELECT%20%22home%22%20FROM%20%22http%22%20where%20time%20%3E%3D%20now%28%29%20-%205m

And it returns this when put in a browser:
{ "results": [ { "statement_id": 0, "series": [ { "name": "http", "columns": [ "time", "home" ], "values": [ [ "2025-06-15T03:10:00Z", 4216.865629946954 ], [ "2025-06-15T03:11:00Z", 2546.421764726118 ], [ "2025-06-15T03:12:00Z", 2308.720535835662 ], [ "2025-06-15T03:13:00Z", 2340.4840379761995 ] ] } ] } ] }

And yes to a Get. But this is what I get when I use the URL in UK

[youzer-name]

Did you select "HTTP(s) - Keyword" as the monitor type?

I don't see the "Keyword: value" on your screenshot below the URL as on mine:

[aesculus]

Well the keyword got me as I was not paying attention. Tried to do the values in the JSON body. Does anything go in there?

Anyway still not working:
http://192.168.1.108:8086/query?pretty=true&db=powerwall&q=SELECT%20%22home%22%20FROM%20%22http%22%20where%20time%20%3E%3D%20now%28%29%20-%205m
Keyword: values

[youzer-name]

I only have the word "values" in the Keyword field that is right below the URL field when editing the monitor. I don't have anything in the JSON body or headers fields.

Have you tried creating a HTTP(s) Keyword monitor and pointing it to some external URL just to confirm that Uptime Kuma is accessing the network ok? You should be able to point one at "https://google.com" with the keyword "Google" and it should show as up.

To test on my side, I created a brand new Uptime Kuma container using the Docker compose file from their instruction page and I was able to set up a new monitor for the database using the LAN IP of my docker host in the URL. I checked my powerwall.yml, influxdb.env, and compose.env files and don't see anything that jumps out at me as different from the defaults, so I don't know why mine is working and yours is not.

The fact that it's working from a browser on another machine means there must be something going on with the inter-container networking on the Docker host, but I'm not sure what else to look at that might be different between our environments.

[aesculus]

Well I tried Google and it worked well. I also tried the internal address of InfluxDB and it failed too.

Its so odd I can get the results via a web browser on my computer but it will not work for UK.

I will poke around a bit and maybe query the NAS (QNAP) forum to see if anyone else has some experience here. NASs tend to be a bit problematic but this area has been pretty normal once they abandoned some of their proprietary stuff last year.

[youzer-name]

Didn't realize you were on a QNAP. I'm not that familiar with them beyond seeing posts here of people having issues with getting Docker set up on them.

My guess is the easiest solution will be to create a separate bridge network and add that to both containers as a second network. Or you could recreate the Uptime Kuma container and put it in the existing PWD network. Once they're in the same docker network you should be able to connect via just the container name or the container's internal IP.

[aesculus]

Well I have played around with it some. I have four other applications on that server including pwpowerwall (port 8675) that use ports (9000, 9001, 8675, 8676) and they all can be monitored. Just not InfluxDB on port 8086.

I will keep hunting.

PS: I wonder how I would monitor pyPowerwall? Its basically just a graphic.