Changing Unifi inform URL via SSH

My house has four Ubiquiti Unifi AP-AC-LR access points. For whatever reason, the construction of the house seems to attenuate WiFi signals significantly from one floor to the other, so I have an LR AP in the cellar, first floor, second floor and in the roof. A while ago I moved my Unifi controller into a Docker container and I changed the IP address with it. Of course, I reconfigured everything, so all the devices used the new Unifi controller.

Fast forward a few weeks. For whatever reason all those access points switched themselves back to the old IP addresses in regard to the inform server. All of them appeared as offline on the device page of the Unifi UI.

In this blog entry aaa.bbb.ccc.ddd is the old Unifi Controller, whereas aaa.bbb.ccc.eee is the new one. The access point is aaa.bbb.ccc.fff.

I tried to fix this by logging into the AP and executing a

    APErdgeschoss-BZ.6.5.62# set-inform http://aaa.bbb.ccc.ddd:8080/inform

But to no avail, the persistent configuration stayed incorrect. The access points were switching to “up to date” for a while and I was able to manage them for a while, but after some time (or rebooting) they switched back to “offline”.

So I checked for the configuration.

    APErdgeschoss-BZ.6.5.62# info
    Model:       UAP-AC-LR
    Version:     6.5.62.14788
    MAC Address: aa:bb:cc:dd:ee:ff
    IP Address:  aaa.bbb.ccc.fff
    Hostname:    APErdgeschoss
    Uptime:      5346 seconds
    NTP:         Synchronized
    Status:      Unreachable (http://aaa.bbb.ccc.ddd:8080/inform)

There was the IP address of the old Unifi controller again and this has been decommissioned for a while (it’s a first gen Cloud Key laying around in my drawer now). So, it was of course unreachable.

When you log into the access point via SSH, you will find a directory /cfg with a file called mgmt.

    APErdgeschoss-BZ.6.5.62# cat mgmt 
    mgmt.is_default=false
    mgmt.led_enabled=true
    mgmt.cfgversion=000000000000000000
    mgmt.authkey=0000000000000000000000000000000000
    mgmt.selfrun_guest_mode=pass
    mgmt.capability=notif,notif-assoc-stat
    mgmt.use_aes_gcm=true
    mgmt.report_crash=true
    mgmt.is_setup_completed=false
    mgmt.servers.1.url=http://aaa.bbb.ccc.ddd:8080/inform
    stun_url=stun://aaa.bbb.ccc.eee/
    mgmt_url=https://aaa.bbb.ccc.eee:8443/manage/site/default

Of course, I tried to change it by editing the file and rebooting everything. But that didn’t work either. As far as I understand the mgmt file is generated at boot from flash.

So, you must actually write the data into this flash before rebooting. I used the following commands after editing the file:

    APErdgeschoss-BZ.6.5.62# syswrapper.sh save-config
    APErdgeschoss-BZ.6.5.62# syswrapper.sh apply-config
    APErdgeschoss-BZ.6.5.62# reboot

The apply-config should suffice. But I’m not perfectly sure of it. And I corrected all four APs before having the idea to try only apply-config.

Nevertheless — after logging into the access point via SSH the configuration was correct and the access point appeared with “up to date” in the list of devices.

    APErdgeschoss-BZ.6.5.62# info
    Model:       UAP-AC-LR
    Version:     6.5.62.14788
    MAC Address: aa:bb:cc:dd:ee:ff
    IP Address:  aaa.bbb.ccc.fff
    Hostname:    APErdgeschoss
    Uptime:      83 seconds
    NTP:         Not synchronized
    Status:      Connected (http://aaa.bbb.ccc.ddd:8080/inform)

I just wasn’t able to recover one of the access points this way. For whatever reason the SSH password didn’t work. I assume an older SSH password was still configured on the device. A password I didn’t have. So, I had to reset it. Of course, this was the one most difficult to reach. In the end I was able to reach it, but with a clear OSHA respectively UVV violation to reach this access point.