On Tue, 2015-09-22 at 10:57 +0200, Pieter Cardoen wrote: > Dear > I am experimenting with dynamic behaviour of NM with multiple networks and > how handover occurs between WiFi/mobile network. I am moving away from the > WiFi access point as far that the NM is unable to connect to the WiFi. NM is > restarted and gets a connection to the mobile connection (lowest autoconnect > priority). Afterwards, I am moving very close to the WiFi AP but NM doesn't > connect to it.
For a specific interface, NM sticks with the existing connection on a device type as long as it's viable, and only when that connection is broken will it cut over to the next best connection. So if NM is connected to a 20% WiFi access point of SSID A but there is an available 90% access point with SSID B, NM will not switch over until the connection to the 20% one is broken. This is not the same as roaming between access points of the same SSID, where wpa_supplicant (under direction of NM) *will* roam between same-SSID access points to achieve the best signal strength. Between interfaces, NM will keep *both* WWAN and WiFi connections active, and switch the default route between the two (and thus the traffic) depending on the connection's "device route metric". So if your policy should be "use WiFi when available", you can set the WiFi device route metric on each WiFi connection to be lower (eg, higher priority) than the WWAN connection, and allow NM to autoconnect WWAN and WiFi. WiFi scanning is also at play here. By default, NM scans periodically starting at 20 seconds and increasing the interval to 2 minutes. This behavior is controversial because scanning can interrupt interactive traffic like SSH or streaming video or gaming. But scanning more frequently obviously updates the list of available access points more frequently too. Given these two goals (better interactivity vs. better information), NM gives you the user the choice by letting you call D-Bus methods on NM to scan more frequently while NM is connected. In the case of restarting NM giving larger signal strength, WiFi signal strength is quite variable even at the same exact location. It depends on interference from other radios/appliances, what TX/RX antenna the WiFi access point is using, what TX/RX antenna *your* device is using, and what driver your wifi interface uses. I wouldn't be surprised at a 10% difference or more, but to get the raw values to look further, run "iw dev <name> link" and look at the dBm values. How much do they change? Dan > If I execute "nmcli dev wifi list" after moving a little bit from the APs > away, I get:* SSID MODE CHAN RATE SIGNAL BARS > SECURITY demo7 Infra 7 54 Mbit/s 22 ▂___ -- > STATION Infra 1 54 Mbit/s 14 ▂___ WPA2 802.1X > CONTROL Infra 11 54 Mbit/s 17 ▂___ -- > forge_wlan_inf4 Infra 64 54 Mbit/s 25 ▂___ -- STATION > Infra 6 54 Mbit/s 19 ▂___ WPA2 802.1X forge_wlan_inf5 > Infra 100 54 Mbit/s 27 ▂___ -- forge_wlan_inf3 Infra 56 > 54 Mbit/s 27 ▂___ -- forge_wlan_adhoc3 Ad-Hoc 56 54 Mbit/s 27 > ▂___ -- forge_wlan_adhoc5 Ad-Hoc 100 54 Mbit/s 24 ▂___ -- > forge_wlan_adhoc5 Ad-Hoc 100 54 Mbit/s 25 ▂___ -- > forge_wlan_adhoc5 Ad-Hoc 100 54 Mbit/s 24 ▂___ -- > This shows me that both station APs (SSID=STATION) have signal of 14% and > 19%. If I am executing "Service network-manager restart" and again ask the > WiFi signal level with "nmcli dev wifi list", I get: > * SSID MODE CHAN RATE SIGNAL BARS SECURITY > forge_wlan_inf4 Infra 64 54 Mbit/s 29 ▂___ -- > forge_wlan_inf5 Infra 100 54 Mbit/s 29 ▂___ -- demo7 > Infra 7 54 Mbit/s 34 ▂▄__ -- forge_wlan_adhoc5 Ad-Hoc > 100 54 Mbit/s 25 ▂___ -- forge_wlan_adhoc3 Ad-Hoc 56 54 > Mbit/s 22 ▂___ -- CONTROL Infra 11 54 Mbit/s 27 > ▂___ -- forge_wlan_inf3 Infra 56 54 Mbit/s 20 ▂___ -- > STATION Infra 1 54 Mbit/s 22 ▂___ WPA2 802.1X* > STATION Infra 6 54 Mbit/s 35 ▂▄__ WPA2 802.1X > My robot is not moved but signal level after NM restart is noticeably higher > than before. It seems that the signal level sticks to the percentages which > have been measured at the beginning of my experiment. > Can someone help me with this issue? > Best regardsPieter Cardoen > > _______________________________________________ networkmanager-list mailing > list [email protected] > https://mail.gnome.org/mailman/listinfo/networkmanager-list _______________________________________________ networkmanager-list mailing list [email protected] https://mail.gnome.org/mailman/listinfo/networkmanager-list
