Upgrade from Powerwall 2 to Powerwall 3

[jasonacox]

Upgrading from Powerwall 2 to Powerwall 3

The Beginning: A Sudden Failure

It began with an alert and a sudden plunge in charge to my Powerwall 2. A firmware update had recently hit. My mind, of course, raced towards that change as the cause. Other Tesla Powerwall users were reporting the same thing. Was this firmware really the cause?

Timeline of Events

• 16 Aug 2025 - System (PW2 and PW+) upgraded to firmware 25.18.5 a411ff15
• 25 Aug 2025 - 19:28 - Powerwall 2 drops offline

Powerwall Charge Graph

The Root Cause: Hardware Defect

No, it turns out that it wasn't the Firmware (though I suspect Tesla needed to push it out to do the rest) the real cause was hardware. A third-party component was defective, and it seems that certain batches of Powerwall 2's were susceptible to this defect (see recall). Tesla proactively sent a command to cause the Powerwall 2 to drain and brick itself. That command was definitely effective! The Powerwall was gone and reported 0 capacity and 0 charge.

Case Created - Service Ticket

27 Aug 2025 - Noticed Tesla App shows “Powerwall Disabled: Service Required - Low Energy Lockout” and submitted request which has been stuck at “gathering parts”

Tesla App Notification

The Tesla App reported “Powerwall Disabled: Service Required" and assured me that service repair would be scheduled soon. But nothing was scheduled.

• 29 Aug 2025 - I sent an email and opened a ticket with Tesla - Response: "We are currently processing your service request and will update you via email on the next steps or outcomes within the next 5 to 7 business days."

• 12 Sep 2025 - Received email from Tesla - Response: "We are replacing a defective Powerwall due to the Low Energy Lockout. We are currently sourcing a replacement and when it is secured from the fulfillment center, we will arrange the site visit to replace it. Once your case has been addressed, you will receive an email or text message outlining the next steps. We kindly ask for your patience as our Tier 2 Department thoroughly reviews the reported issue."

• 20 Sep 2025 - Still waiting - Tesla App now says "In Review"

We are now a month+ after the Powerwall failed and I'm a bit worried that I got lost in the ether. I decided to call.

• 27 Sep 2025 - Called Support at (877) 961-7652 option 2 - Design team is reviewing system to see if they can upgrade it to Powerwall 3

I wasn't expecting an upgrade! I expected they would replace the defective Powerwall 2 and be done. The technician explained that they would replace the old system, the Powerwall + (basically a Powerwall 2 with an inverter) and the old Powerwall. They would somehow use the old Gateway 2 since it was able to separate the Powerwall from the grid, allowing for a whole-house backup that was one of the key reasons I decided on Tesla to begin with.

Scheduling and Delivery

• 3 Oct 2025 - I received notification in the Tesla App that service needs to be scheduled. First date available was 15 Dec 2025, so I picked that. Install would only take 2 hours.

• 11 Oct 2025 - I get an email response from Tesla: "Powerwall replacements are currently taking anywhere from 6-8 weeks at this time. Your ticket is currently pending review from our design team but should be moving along soon."

• 8 Dec 2025 and 14 Dec 2025 - I get an email reminding me of the pending appointment on 15 Dec 2025.

• 12 Dec 2025 - On Friday afternoon, a surprise knock at the door revealed a semi-truck parked in the street in front of our house. It was Tesla delivering the new equipment for the service repair. It was a huge 5' tall pallet, and they asked to drop it off in the driveway. Fun fact: We had a scheduled Christmas party that would be happening at our house with dozens of families. We managed to tuck the huge pallet in our garage and turned it into a decoration and novel discussion piece. It worked out for us, but I'm surprised that Tesla didn't have some way of notifying us to be home to receive a large package like this.

Powerwall 2 system with pallet of new gear

Installation Day: 15 Dec 2025

• 15 Dec 2025 - Installation day! The notice said to expect the Tesla team at 7am-1pm. So, of course, they arrived at 1pm. To be fair to them, traffic was terrible. But the team was amazing. The two went right to work removing the old Powerwall system.

I met the techs in the driveway and helped them orient to what we had. I had asked the call center if I could get a copy of the new design documents, so I asked the tech. He informed me that he didn't have them either. I asked if that was common, and he said they have the docs, but he doesn't have them and will just have to call in to get details. But he assured me he had done many of these and knew what to do. I'm not worried, but I am disappointed. I was hoping to have the diagrams.

Saying Goodbye to Powerwall 2

I really liked the look of the old Powerwall 2 systems. Their aesthetic design is much better than the clunky Powerwall 3 metal boxes that really don't look all that different from any utility box other than the glowing logo. The old Powerwalls were put in ship-back boxes on the old pallet. I know it was dead, but I was still sad to see them go. The technician did agree with me that the old design had much nicer trim, and the glowing pulsating green bar was definitely a high-tech look. Goodbye Powerwall 2!

The new Powerwall 3 units were identical. Unlike the Powerwall + that only supported 4 solar panel strings, each of the Powerwall 3 units had an inverter capable of supporting 6 strings. I asked if that means I could add more panels. Of course, but they said they couldn't do it today! Ha!

Pro Tips for Installation

Important: Document the QR Code

As the installers are mounting these new Powerwall 3's, the front cover will be open. Make sure you take a picture of this and especially the QR code directly below the Tesla logo backlight. This has the SSID and the Password you will need to join to the access point. This password is also used for the TEDAPI connection.

If you don't manage to get this before the cover is on, you will need to remove the four screws (2 on each side) to slide it off.

I have two Powerwall 3's. There will be one that is the primary. For me, it is the one where all the solar array strings terminate. This is the one I used to connect to for the Powerwall Dashboard.

The Installation Process

It took more than 2 hours, but I sure didn't mind. It was interesting to see the ingenuity that the techs had to apply. The Tesla "lift" that they used to hang the Powerwalls would not go high enough. They had built a makeshift platform that they had to use to raise it far enough.

I love these kinds of things. The engineers back at Tesla HQ probably had no idea how their nifty lift doesn't do the whole job, and each of the tech teams have to invent a solution. Of course, the solution took a lot more manual labor and extra time. I'm convinced that engineers need to be sent out to the field a lot more to shadow their techs to see what needs to be improved. I digress.

For my setup, I have two strings. But one of them is a combined string (essentially 2 in one). This requires that you bridge those strings across two of the inverter inputs. This almost didn't happen, but one of the techs noticed the gauge of the wire was different and told the other that they needed to find the bridge. Thankfully, they had those and did get them installed.

Wired Ethernet

When the installer was hooking up my Powerwall 3's, and asked for my WiFi password, I asked if we could hardware them instead. He said yes and was so relieved! He said they see a lot of problems with WiFi (yes, we all know) and having a hardline will make upgrades and operations better. I had a CAT6 line to the old units so I just crimped a new end on it and let him plug it in. He used another CAT6 line to connect the two Powerwall 3's together (fun fact: there are two ethernet ports on each Powerwall for exactly this reason).

Recommendation: Have the installers hardwire your system into your LAN if you can. From what I learned, using WiFi to do the installation would have easily taken up another hour to download the firmware. The reliability of hardwired ethernet is so much higher than WiFi, even with those "made by Jason" cables.

Post-Installation

Star Wars Crates

The old Powerwalls were put in boxes and a fancy Star Wars looking crate, ready to be picked up by the Tesla team. Those would eventually be picked up in 3 days.

Online - Powerwall Dashboard Back

So far so good! It was evening so there wasn't any solar production. I noticed the Tesla App was working again, but still had the warning about the Powerwall Disabled. That warning would go away in a few days.

I logged into my Powerwall Dashboard host and switched the WiFi to connect to the new Powerwall:

# Connect WiFi to new Powerwall 3 access point
sudo nmcli device wifi connect "TeslaPW_ABCDEF" password "XYZABCXYZ" # some may require adding: ifname wlan0

It connected. I could ping 192.168.91.1 without any problems.

I then ran ./setup.sh, selected "4 - Extended Metrics," and told the system to replace all previous settings. I entered the new TEDAPI password (same as WiFi one above), which I had saved from the QR code from the primary Powerwall. I loaded the Grafana dashboard. It worked!

After a day, zero connection errors. I know there are a lot of you who have had issues getting solid connections to your Powerwall 3. So far, my setup has been stable. I also added a Raspberry Pi (similar to setup here: #607), and it connected as well.

Setup Glitches

The Powerwall Dashboard was working perfectly. However, I noticed a few things that were still off. The system was not exporting to the grid, and PTO was not selected in the app. Also, the system was not able to charge from the grid (e.g., storm event or other reason to charge). I also noticed it had the wrong utility rate plan under time-based control. I figured the tech probably didn't know what to set since he was operating without any plans.

So what do I do? Well, I could put in a call and wait until next Christmas, or I could just fix it. At this point, I should say that I wouldn't recommend this for everyone. Proceed at your own risk, and with caution, if you decide to do the same! :-)

Using Tesla One to Fix Configuration

I fired up the Tesla One app and connected to the Powerwall. I scrolled through the settings. There it was! It was set to disable all grid charging, and the utility setting was incorrect. I fixed those and switched back to the Tesla App. Yep, everything was as it was before!

Conclusion

So far, I have been extremely happy with the replacement. The Powerwall 3's are doing well. However, I did notice some things with pypowerwall and the Powerwall Dashboard that had me concerned. For the first thing, the memory usage was skyrocketing and indicating a memory leak. It seems the container may even be restarting occasionally.

Image of pypowerwall memory usage over time

Optimization

As many of you know, it has been difficult for me to help you with your Powerwall 3 problems. The Powerwall 2 behaved so differently. Thanks to the work from several of you in the community, we managed to maintain functionality. But it definitely made it difficult for me to test out any changes or improvements I was thinking of for the Powerwall 3.

I started looking at the calls and noticed some of the calls were taking nearly 2 seconds to render. I have a lot of automation, including other displays that use pypowerwall to render status. This latency was noticeable and something I had not seen with the Powerwall 2. I decided to spend a day addressing these latency issues. It seemed to be coming from the TEDAPI section of pypowerwall. I have been seeing issues reporting this, and the fix was always to increase the cache expiration time. I did that and it helped, but I wanted to see if there were other things we could do to keep the 5-second default working.

I went to work and identified several areas that could be improved. These are all captured in this pull request: jasonacox/pypowerwall#247

Performance Improvements

Running tests with the new perf_test.py tool, comparing before and after (reflecting new cache impact):

API Route	Before (ms)	After (ms)	Improvement	Usage Count	Impact Reduction
/api/meters/aggregates	821.5	151.3	81.6% ⚡	7,992	5,355 seconds saved
/aggregates	764.8	150.7	80.3% ⚡	3,880	2,383 seconds saved
/strings	545.7	37.3	93.2% ⚡	3,945	2,006 seconds saved
/vitals	339.3	33.9	90.0% ⚡	3,946	1,205 seconds saved
/alerts/pw	382.9	87.8	77.1% 🚀	3,881	1,145 seconds saved
/temps/pw	266.0	253.6	4.7% ✅	3,880	48 seconds saved

Updated test image used for these: jasonacox/pypowerwall:0.14.5t86-beta5

Also, after the changes, I'm seeing much better and more stable memory management:

Image of Graph

I have a few more ideas I want to try. I will get this new release in place for an easy upgrade.

Note to Powerwall 2 Users

If you are still a Powerwall 2 or + user, I need your help. While I'm focused on improving the experience for Powerwall 3 users (and enjoying that improvement myself), I'm keen to make sure we keep the software working well with the fleet of Powerwall 2 installations. Please make sure you raise any issues.

[SCHibbard]

Man, quite the saga, Jason! You'd think the Tesla Marketing team would keep an eye out for you, considering the indirect free promotion you give them. I've got 2x Powerwall 2+. Been wondering about adding one or two, but struggle if I should try to find 2 more 2+ units, or go all 3s. Maybe if I wait I'll suffer the same fate and can get all 3s. In the meantime I'll keep an eye out on PWDB usage with 2s!
---Scott---

[jasonacox]

Thanks @SCHibbard ! 😁

Fun fact: When I called the Tesla support team, I asked if I could request just a Powerwall 2 replacement. Part of my logic was that a Powerwall 2 upgrade should be easier (swap out) and might be able to move up the install date. As I said, I also liked the Powerwall 2 design and the fact that it still had LAN accessible APIs. The support rep told me that he would add the note to the ticket but he though the upgrade was likely going to happen. When they arrived, I asked the techs about the Powerwall 2's and they said they were no longer in production so all replacements and new installs were Powerwall 3's. I'm definitely glad now that I got the upgrade. It is going to be much easier to support this community having the latest gear to troubleshoot (almost all issues now are PW3 related).

For your case, I suspect some third parties may still have inventory for the Powerwall 2's. Or, as you say, check to see if you are on the recall list. The details I found:

• The recall specifically targets units manufactured and sold during a nearly two-year window:
• Production/Sale Dates: November 2020 through December 2022.
• Geographic Scope: Recalls have been issued in the United States (approx. 10,500 units) and Australia.
• Identifying Labels: Affected systems have the model number "Powerwall 2" printed on the nameplate label on the side of the unit.
• Specific Part Numbers (Reported): Some users have identified affected part numbers, including 3012170-05-A, 3012170-05-B, 3012170-10-B, and 3012170-25-B.

[SCHibbard]

Thanks, @jasonacox. Mine were installed 2023-02, so they COULD have been in the recall, but the TPNs are both 3012170-05-E. Looks like the recall only included -A & -B. Oh well!

[JohnJ9ml]

Thank you for the performance improvements. They've been bothersome, but have been hard to convey, and I wasn't sure that this was normal behavior. I'm testing them out now.

[jasonacox]

Please let me know what you find, @JohnJ9ml 🙏

[JohnJ9ml]

I have been running pretty consistently for the last five days, and do the following:

a call to power() every five seconds.
a call to strings() every ten seconds
a call to a pile of endpoints (including vitals, battery blocks, and aggregates) every 30 seconds
a call to alerts/pw every minute.

For the most part, everything holds up fine. Certainly, the proxy holds up fine. I still see the timeouts, and I still enter these peculiar states where fetching the level() works, but nothing else returns data. I know that's on the PW side, not the proxy side, but there are times I wonder if I'm overwhelming the limited capabilities of the PW3 CPU. Purely a hunch on my part.

I have three PW3s all communicating back and forth, and there are times I believe they just don't prioritize those ports or something, because I feel that I saturate the processing by abusing it this strongly.

There is something not quite right about my configuration in that it tends to forget how it's configured. I have a call in to my installer, but this is something I've been tracking over time and have enough information that I feel I have the general rhythm of the system. I have strings feeding each of the three inverters, and occasionally one or more of the inverters lose track of the fact they have source feeds coming to them, and stop contributing to my installation. The only recourse is to turn the unit off for an hour and then bring it back up. This can happen to any of them, including the lead PW3. (which is interesting to have powered off the primary unit and continue to get data from the installation. About like rebooting your car while driving (show of hands - how many have done that? I have).

There are times I've seriously wondered how much load monitoring it puts on the system. I note the energy loss through the night that seems to exceed what the house is drawing. Makes sense it takes power to run the unit, but I wonder how much extra is involved just to respond with the data probes being performed.

The things you can think about with a lot of data around as fodder.

[jasonacox]

A year ago I introduced aggressive "retry" logic in pypowerwall that would hammer the Powerwall API for "vitals" data. It turns out that it would even do this for 404 response. When Tesla removed the vitals endpoint and sent back constant 404s, it caused a non-stop avalanche of calls. I had not seen this yet (my Firmware had not upgraded) but I heard from a friendly Tesla engineer who suggested I change the logic. They were seeing logs and hearing from customers about this fault. He assured me that it wasn't critical because the gateway is designed to ignore and rate limit aggressive traffic, and the APIs are isolated from the actual system operation, so that it wouldn't impact the operations of systems (e.g. the CPU overwhelmed hypothesis you mention). Still, this has been a big concern of mine and I was relieved to hear this protective design principle in the system architecture. I don' tknow if that still applies to the new Powerwall 3 systems, but I suspect it does.

All that to say, while you seem to sense a correlation, it does not seem likely that the monitoring traffic you are sending is causing the operational issues. And, if so, it is bizarre because the traffic you mention below is very low. I am doing at least 10x the number of calls at a much higher frequency with no issues with very low error rates (sub 0.1% via /health reports).

Suggestions:

• Turn off active / local monitoring. I would suggest running ./setup.sh and switching to Cloud Mode. It would not put any load on the Powerwall yet you would still have more trend data you can use to help isolate issues. Yes, the Tesla app could give you some of that but it is a bit more difficult to navigate / compare /extract.
• When it locks up again, if you stop monitoring, does it ever recover by itself? If it doesn't recover until next morning (sunrise), I would suspect the MPPTs or the Inverter itself.
• When the installer was hooking up my Powerwall 3's, and asked for my WiFi password, I asked if we could hardware them instead. He was so relieved! He said they see a lot of problems with WiFi (yes, we all know) and having a hardline will make upgrades and operations better. I had a CAT6 line to the old units so I just crimped a new end on it and let him plug it in. He used another CAT6 line to connect the two Powerwall 3's together (there are two ethernet ports on each Powerwall). I don't know how they installed your setup, but I would explore if the Powerwall's are communicating with each other via WiFi or have a hard line connection. Ideally they would use a hardline connection between Powerwalls as well as a hardline connection to your LAN and not use WiFi for internet access. I actually think this may be why my TEDAPI access has been so solid. The Powerwall access point is only serving my requests, not trying to communicate to the Tesla Cloud via the same WiFi bandwidth. But then again, others in the community have tried that without the same results.

I hope that help. Let us know if you learn anything!

PS - Happy New Year! 😁

[JohnJ9ml]

This was purely my musings from just sitting and watching the whole thing run. I have a bad habit of just sitting and watching systems like this run and observing the ebb and flow to the systems, and then letting my imagination run away.

I don't have the PW dashboard running, so there is still only me hitting up the traffic (dashboard stopped running when I did an update, been waiting for a time to devote to getting it set up again).

I know my units are hardwired together (the two ethernet ports get used to daisy chain between multiple PW3s) and there is also cat6 running to the PW that is being used for comms. I wish the WiFi wasn't even configured on the PW, but I know it's belt and suspenders approach to calling back to the mother ship.

Anyway, I was just thinking about what seemed to be happening and wondering about the load.

If you're caching every five seconds but say you're making 10x the calls I am, are you caching a lower setting or getting something else out of the polling? I was happy to return to five second caching because it better smooths out the assumptions on the "accumulator" values and trying to extrapolate them over the course of the day. I'm hesitant to ratchet things up any higher, but your 10x load sounds... interesting.

and Happy New Year to the whole community here.

[jasonacox]

In my setup, because I usually have several tests running, I often have 5 different systems running the Powerwall Dashboard stack separately, all polling every 5 seconds for /aggregates, /vitals, /strings, /alerts, etc. They are all using a separate 5s cache, so the hits are combining. However, I only have 2 Powerwall 3's now (versus your 3). I did see rate limiting on the Powerwall 2's, but haven't seen that on the 3's. Of course, that is one of the main reasons I added caching to pypowrwall. Anyway, your observations could be right, but it would be disappointing.

[anderep]

Mine eight powerwalls were installed in December 13 2022 and are -E parts. They have been doing good so far with capacity at 14.71 - 14.33 kWh.

[mcbirse]

Thanks @jasonacox for sharing your upgrade experience!

I've finally read through your post, having been away for a few weeks - trying to catch up on posts! Wow, I go away for a few weeks and come back on we are now on Grafana v12.3! Jason, you need more holidays. 😄

Anyway, glad to hear your upgrade went well, with only a few minor hiccups with the system setup (that you were able to resolve yourself).

Four months on from my upgrade, and I can't say the same, unfortunately. 😡

Hoping that maybe from looking at the setup of your system with Tesla One I might be able to resolve my issues. I'm about to give up on Tesla Support fixing the problems after chasing them for 4 months!!

My initial install experience was mixed - Tesla Support was great, but my installer was not (and I'd rather not have to ever deal with them again):

• Tesla scheduled the replacement within a reasonable timeframe and delivered to the installer within their estimates.
• However, not hearing anything a week or so past the due install date, I contacted Tesla for an update to be told my PW3 unit had already been delivered to the installer the week before.
• I then ended up chasing the installer, who refuted what Tesla was saying. Eventually I got more info from Tesla directly, including the date/time my unit was delivered and the name of the person at the installers address who signed for it.
• Only when my installer was presented with this information, was my PW3 mysteriously "found" in their warehouse. They were not apologetic at all, rather they were rude and arrogant - hence why I never want to deal with them again. Not a good start to getting my system replaced....
• The installer then came to site promptly at least, and replaced my PW2 with the PW3 which took a few hours.
• They advised mine was the first replacement they had done - which makes me wonder if they configured the system incorrectly, resulting in the problems I'm now seeing in the Tesla App.

After installation of the Powerwall 3, I quickly got Powerwall Dashboard setup again - everything connected vi WiFi and TEDAPI without issues. ✅

I did note similar issues to yourself however, with PTO disabled, grid charging disabled, wrong solar setup (solar roof, not panels), etc. which I fixed myself via the Tesla One app.

Monitoring the Powerwall 3 via Powerwall Dashboard has been flawless, and is working perfectly and showing all the correct information. 🎉

However, since the replacement, I am seeing completely wrong information in the Tesla App along with a couple of other issues:

• Tesla App Energy values are showing doubled values, i.e. solar generation, home usage, export, is double the actual amount (and I can't use the tesla-history script, as it also returns doubled values).
• Going off-grid fails. When selecting Go Off-Grid in the Tesla App, my system will briefly go off-grid, then automatically return to on-grid. It will not stay off-grid.
• The Backup Test via Tesla One app was initially also failing, reporting "One or more Powerwalls have problems". This has since started working after receiving firmware update to 25.34.1

Below is an example of the doubled energy values in the Tesla app:

That's a hell of a lot for a 10kW system! And almost exactly double what I see in Powerwall Dashboard.

I have been in contact with Tesla Support numerous times in the last four months, trying to get this resolved. Despite providing them extensive information, they have still failed to provide any resolution which is beyond frustrating.

Trying to diagnose the possible cause myself, one thing I have noted is when I connect to the Backup Gateway via Tesla One (TEG WiFi, not TeslaPW), I can still see the old Powerwall 2 system (and errors relating to this),

It also shows live meter readings (maybe that's normal). My speculation though is these live meter readings from the Gateway are being aggregated to the meter readings recorded by the Powerwall 3 - hence the doubling of values in the Tesla App.

@jasonacox - couple of questions if you don't mind checking your system?

• What do you see if you connect to the Gateway via Tesla One? (TEG WiFi)
• Did Tesla add your Powerwall 3 system to the existing site in the Tesla App (they did with mine), or is it a different/new site in the app? (different site id)

Tesla Support has been no help. I really want to get these issued resolved - any suggestions?

[jasonacox]

Jason, you need more holidays. 😄

Haha! Yes! But I confess, I was addicted to vibe coding over the holidays, which helped. The move to Grafana 12 would not have happened without it. I pointed the AI at the Grafana code base and asked to help me make the changes I felt were important for a seamless transition (e.g. centered panel titles, replacement panels) and had it update the power flow animation code to better match the color scheme. Yes, minor things but only 10 min. for our friendly neighborhood 🤖. :)

Four months on from my upgrade, and I can't say the same, unfortunately. 😡

Oh my goodness, @mcbirse !! This is terrible!

I have the fortune of having Tesla be the OEM as well as the installer. Watching reddit threads as well as your experience, I have come to understand how important this is. Also, sadly, Tesla is not in every market for installation. Lesson to anyone listening, try to get Tesla to be your installer if you can, otherwise spend extra time trying to find the best installer. The experience you had with your installer is disappointing and massively unacceptable.

I'll make a few updates to this, but first pass:

What do you see if you connect to the Gateway via Tesla One? (TEG WiFi)

I don't know how they did it, but my old TEG access point is gone completely. I suspect they disabled it on purpose. I'll do some more research because the Gateway (GW2) is still being used. I do know they removed my hardwired ethernet (as mentioned in my post above) which I recrimped to plug in to the PW3. But that wouldn't be the access point. They did make changes to the GW2 physically too, but I wasn't able to see all of that. I'll investigate.

Did Tesla add your Powerwall 3 system to the existing site in the Tesla App (they did with mine), or is it a different/new site in the app? (different site id)

Same site ID. They did mention it would take a few days to see the old Powerwall 2 details disappear. It did. However, I'll poke around a bit more in the Tesla One app to see what else I can find.

Tesla Support has been no help. I really want to get these issued resolved - any suggestions?

I asked Tesla Support to give me the new system diagram PDFs. They did. It shows what was changed physically. I'll look over it a bit more to see if there are things you should try. Also, have Tesla Support send you those plans too if you haven't already.

[mcbirse]

Thanks Jason. The fact you cannot see the old TEG access point is... interesting. I'm convinced my installer did not follow the correct setup/replacement procedure. It feels like my GW2 is still in the old configuration, which it shouldn't be. I have another call in with Tesla Support and will see how that goes.

[mcbirse]

Hooray - finally, I can report that all issues with my system have been resolved! 🎉

I was on the verge of performing a "Factory Reset" via Tesla One due to frustration and setting everything up from scratch. Pretty sure this would have fixed it, considering what Tesla ended up doing...

So, after more back and forth with Tesla this week, the engineering team resolved the issues by "recommissioning" my system. This was able to be done remotely - no site visits required at all (why did it take over 4 months for them to do this!? 😠 ).

From what I can tell, they performed a factory reset to recommission the Powerwall 3 (with Gateway 2) from scratch.

There were some minor settings not configured correctly afterwards (same as mentioned previously), which I fixed via Tesla One, but otherwise, everything is now working correctly.

• Selecting Go Off-Grid from the Tesla app now functions correctly. The system stays off-grid until I select to return to grid.
• Performing a Backup Test in Tesla One passes with all tests successful.
• The system operates correctly when simulating a power outage (provides power to my home while also still allowing solar production/charging from solar).
• The Tesla app is now showing correct historical energy values (i.e. daily solar, home, powerwall, and grid energy usage/production - all correct).
• The tesla-history script is working properly for me again. 😉

Of note, after the system was recommissioned, the "TEG" access point of the GW2 is no longer visible.

This appears to be how things should be, given @jasonacox mentioned it as well. If anyone is seeing the "TEG" access point with a PW3 + GW2 configuration/replacement, it’s possible there is something wrong like there was with mine. Only the "TeslaPW" access point should be visible, I believe.

I suspect the commissioning process (i.e., if done via Tesla One, by connecting to the PW3 "TeslaPW" access point, not the GW) will reconfigure the gateway such that the "TEG" access point is disabled.

Also, like @jasonacox, my installer had removed the direct ethernet connection to the GW2 (moved it to the PW3).

With the reconfiguration and comms disabled, reporting of any meter readings by the GW2 back to Tesla won’t happen (this is done by the PW3), which is what I think was causing doubling of energy values in the Tesla app. The other issues I had with backup tests failing and Go Off-Grid not working were likely due to the failed PW2 still being registered in the gateway.

How’d it all go wrong?

Well - I mostly blame my installer. I don’t think they followed the correct process for replacing my PW2 with the PW3.

The commissioning process should set everything up correctly, including reconfiguring the GW2 as required. From what I have read in Tesla documentation, one important step seems to be ensuring the full device setup is run (not simply scanning for new devices).

The other possibility is that the gateway required a firmware update before connecting and commissioning the PW3, and my installer did not do that.

Either way, if my installer had simply tried to run a Backup Test from within Tesla One afterwards, they would have known things were not right immediately!

My experience afterwards with Tesla support trying to get the issues resolved has been just as frustrating. This was a warranty job in the first instance, so I couldn’t exactly call the installer (nor did I want to call THAT installer, for reasons I noted earlier).

It’s taken over 4 months since the replacement for Tesla to resolve the issues. Really not good enough, in my opinion.

[jasonacox]

Wow! First of all, congratulations @mcbirse for the reward of your persistence! Glad it is working again. Also great discoveries. Hopefully that will help others. As you say, still a terrible expereince.

The tesla-history script is working properly for me again. 😉

And the entire community is happy for that too! 🙏 Thanks for contributing and maintaining that great tool. Now with a working system again. 😆