Powerwall-Monitor: Is this normal?

[komrade]

Running the Grafana dashboard and noticed that the data stream is occasionally offline/no data for few minutes everyday. The Gateway is connected via the ethernet. WIFI is off.

See the following image. Is this normal?

Thanks!

 

[cwied]

No, this doesn't look normal. The outages indicate either telegraf went down or lost connectivity to the Powerwall. I only see gaps like this when either the Powerwalls updated firmware or I did maintenance on the machine running telegraf and influxdb. You might want to check the logs on the server running telegraf and/or influxdb. I eventually moved Grafana onto a different server because when they were all on the same server, it started running out of memory.

Here's a typical graph for me:

 

[komrade]

No, this doesn't look normal. The outages indicate either telegraf went down or lost connectivity to the Powerwall. I only see gaps like this when either the Powerwalls updated firmware or I did maintenance on the machine running telegraf and influxdb. You might want to check the logs on the server running telegraf and/or influxdb. I eventually moved Grafana onto a different server because when they were all on the same server, it started running out of memory.

Here's a typical graph for me:

View attachment 811999

I see. I'm running the instance on GCP. What's the recommend memory allocation for Grafana/InfluxDB?

PS: Is there anyway to check the logs of Tesla Gateway to see if the connectivity is being dropped?

 

[cwied]

I see. I'm running the instance on GCP. What's the recommend memory allocation for Grafana/InfluxDB?

PS: Is there anyway to check the logs of Tesla Gateway to see if the connectivity is being dropped?

I don't know of any way to get to internal logs on the Tesla Gateway, unfortunately.

Are you using InfluxDB cloud or running everything on a VM? As far as I could find in my research, InfluxDB grabs all the memory of the server it's using, so it's best to dedicate an instance to it. Note that I'm running InfluxDB 1.8. I don't know if the 2.x versions are any friendlier. I'm running both InfluxDB and Telegraf on the same VM and that seems to be working ok, but with Grafana running there too I was seeing processes getting killed occasionally by the OOM Killer. That might look the same as your symptoms.

 

[komrade]

I don't know of any way to get to internal logs on the Tesla Gateway, unfortunately.

Are you using InfluxDB cloud or running everything on a VM? As far as I could find in my research, InfluxDB grabs all the memory of the server it's using, so it's best to dedicate an instance to it. Note that I'm running InfluxDB 1.8. I don't know if the 2.x versions are any friendlier. I'm running both InfluxDB and Telegraf on the same VM and that seems to be working ok, but with Grafana running there too I was seeing processes getting killed occasionally by the OOM Killer. That might look the same as your symptoms.

I'm running everything on a VM. I wasn't aware that InfluxDB will grab all the available memory. Most likely that's what causing the issue.
What's the decent cloud/hardware solution to run this on? GCP free tier clearly isn't worth it.

 

[cwied]

I'm running everything on a VM. I wasn't aware that InfluxDB will grab all the available memory. Most likely that's what causing the issue.
What's the decent cloud/hardware solution to run this on? GCP free tier clearly isn't worth it.

I don't think you actually need a lot of resource for influxdb, you just need to dedicate the whole vm to it. I have it running on a single core VM with 2GB of memory allocated on an old FreeNAS box (it's only an i3-4130). I'm running Grafana on a separate mini-PC that also runs OPNsense and Home Assistant.

If you can somehow get another free tier service elsewhere to run Grafana, you might be able to still make it work in the cloud.