Client Connectivity Quirks in Omada Wireless Network
Problem Statement:
- WiFi distribution network over 9 AP's works fine 95% of the time. The other 5% is intermittent enough that makes it difficult to attribute cause.
- Client (smartphone, or TV streaming device) can go into buffering, and streaming stops. Slow internet browsing. Speed can drop to modem speeds (56kbps as an example).
- Attaching to a different SSID or cycling device power seems to restore full connectivity. But issue will happen again in the next 20m to 2 hours.
- With few clients on this network and light traffic, SL speeds seem to be typical at first connection (80 Mbps+). Eventually, the DL speed can be artificially slow (5Mbps DL, 13 Mbps UL), that show up as a faster upload. Issue continues to degrade towards dial-up modem speeds. Switching to SL directly (phyisically) can yield the 80 Mbps+ speeds, so SL speed is ruled out.
- Putting an independent WiFi router on a SL terminal yields maximum SL speeds with no loss of connectivity or buffering issues.
- MOOCH is the guest network, open and BW limited. This issue can affect this too.
- Logs show occassional failed authentication for a client device as they roam from AP to AP. Following two examples are the only type of failure events logged.
- [Failed]Joe-s-S22-Ultra failed to connected to Bldg2 2nd Flr with SSID "Bldg2" on channel 11 because WPA Authentication failed.(1 time in a minute)
- [Failed]joe-s-S22-Ultra failed to connected to Bldg2East with SSID "Bldg2" on channel 6 because the Association times out.(1 time in a minute)
- No other logs seem to show any useful clues. Seems I'd have to use a laptop until the issue is experienced and do a Wireshark capture.
- Not dependent on signal strength. Issue can happen with -70 dBm 5 GHz or 2.4 GHz.
- Seen "No Internet Connectivity" issues on smartphones. DHCP successful, but connection stated as "no internet". Client devices successfully responds to PING. Older phone, or other compatibility issue not ruled out. Could be red-herring.
- Overall, the clients are happy, and don't seem to notice issues during sporadic use. Streaming failures are noticed and complaints made.
WDS description:
Omada 9 AP network covers 3 buildings and surrounding area. Three Omada switches, GW and OC200 controller. Internet is sourced from two SL terminals located on different buildings. While one SL is colocated with GW and Switch, the second is on a different building and brought back to the GW through as VLAN30 over Ethernet, then untagged and fed through an Ethernet SFP into the SFP WAN GW port. VLAN30 sole purpose is a point-to-point transport to bring the second SL internet feed to the GW.
LAN is default for "Building 2" SSID and is also the management LAN. This means 192.168.0.0/24 is the management LAN and the client facing distribution LAN.
Building 3 SSID operates over VLAN33 and is 192.168.33.0/24. Same issues.
Configuration options:
Load Balance is enabled.
SSID routing is enabled to send each SSID through it's respective SL terminal. Easy to turn off and revert to only load balance. Both methods attempted, with same results.
I recall that even with only one SSID and one SL terminal, this issue did happen on the default LAN (management and client).
Have NOT attempted to put both SSID on VLAN and separate management... leaving it on default 192.168.0.0/24.
Firmware for ALL devices is up-to-date relying on Omada SDN update feature.
This is like some sort of memory leak...or the GW looses track of client traffic switching. What next? Where to look? Any insight or vectors are appreciated.
See diagram.