The problem persists. Not all Google Wifi users are effected, but now the issue has occurred at our offices so I was able to get a packet capture. We have 2x Google Wifi units on software version 11895.108.11. Now using ESP-IDF v3.2.
The one thing that looks NOT RIGHT to me is, the ESP32 initial tries to communicate with a different Google MAC than the 2x MACs that I see on the successful connections. GFB rather than GF5/G80. Not sure if that is relevant.
I'd appreciate any insight as to whether others have experienced compatibility issues with Google Wifi!
Next I intend to try upgrading to ESP-IDF v3.3.
----
Here's a ZIP with 3 different packet captures and ESP32 serial logs:
https://bond-updates.s3.amazonaws.com/m ... 2_pcap.zip
The captures are labeled:
- fail = we did not establish a connection at all
- stumble = we did eventually get connected, but it took a while / unstable
- stumble2 = similar to above
Here's a summary of what happens in the "stumble2" capture, where it takes ~12 seconds to get connected rather than the nearly instant connection that we experience with ESP32 in other networks:
Code: Select all
Shorthands for devices:
ESP32 = 3c:6a:2c:10:00:94
GF5 = Google_61:56:f5
G80 = Google_61:27:80
GFB = Google_61:56:fb
Summary of "Stumble2" capture:
Time 0.0s: ESP32 sends 3x Prob Requests for "Googletest" SSID, at 120 ms interval
Time 5.5s: GF5 starts sending QoS Data, and G80 starts sending Request-to-send to ESP32. But no response from ESP32.
Time 5.6s: Now GF5 stops sending. G80 now sending 802.11 Block Ack Req. Continues until 6.1s.
Time 12.6s: ESP32 sends Dissociate to GFB (!), with 10 re-transmissions within ~4ms.
Then, in rapid succession (<100ms total):
ESP32 sends Authentication to G80.
G80 sends Acknowledgement (?) and then Authentication to ESP32.
BOND sends Association Request to G80.
G80 sends Acknowledgement (?) and then Association Response to ESP32.
A key exchange procedes.
It appears we are now connected.