Hello, I have been struggling with the problem of access points falling asleep after a long time without clients connected for a long time.
Configuration quite simple “Using CAPsMAN Forwarding Mode”. Everything seems to work properly where there are constantly connected clients, while the APs to which nothing is connected stop working for a few hours. ie they broadcast SSIDs, we can connect to them, but we will not get an IP address anymore. As if they fell from the bridge onto the capsman. After disabling and enabling the capsman interface, everything returns to normal.
Interestingly, after restarting the capsman interfaces, all these interfaces show up as dynamic in bridge> port, and that’s correct. However, after a while, these dynamic ports disappear, and where clients are connected, everything still works. And if he wants to use an AP that no one was connected to, it does not work anymore, no IP because the interface is not in the bridge.
I don’t have experience with such behaviour under CAPsMAN. With stand-alone wlan interfaces it’s normal that they become inactive (i.e. without R flag) when no client is registered. This is normally not a problem, interface changes state to “running” quickly enough for initial traffic between wireless client and the rest of network (most likely that’s DHCP lease request) pass without a problem. However this can be a problem if bridge (to which wlan interface is attached) runs any of xSTP because xSTP causes a delay when port becomes active to test for any loops. Delay can be long enough for wireless client to freak out.
This behaviour can be disabled by setting disable-running-check=yes on wlan interface which makes the interface to appear running all the time. However there’s no such setting in CAPsMAN. Some settings which can’t be set via CAPsMAN provisioning are inherited from on-device setup, you could try to disable CAPsMAN client of CAP, set this option and re-enable CAPsMAN client to see if behaviour changes.
I was interested in this too, and the wireless interfaces dynamically added to the bridges by CAPsMAN are marked with edge=yes, so no forwarding delay should take place after they become active.
But I confirm that disappearance of CAPsMAN-controlled wireless interfaces from the /interface bridge port is not a normal behaviour.
So @darkneess, what is the RouterOS version and device model on which you suffer from this?
CAPsMAN Controller/Router: RB3011 - 6.48.1
Switch: CRS326 - 6.48.1
CAP - RB931, RB941, RB951, RBD52G and other - 6.48.1
this is what it looks like after restarting the router or all interfaces in “Cap Interface”:
after a few hours it looks like this
Everything still works wherever the clients were connected. However, where there were no customers, each newly connected is not in “bridge”
We are aware of an issue where the CAP interface wasn’t added to the bridge, the fix is available in versions after 6.49beta16. So you shouldn’t notice this problem in the beta branch. But you can write us a support ticket and we can double-check if it’s the same issue.
It works!
In version 6.49beta27 the problem does not occur, everything works as it should without interruptions and disappearing interfaces from the bridge. So I’m waiting for the stable version and for now I’m using the beta.
Interestingly, I remembered that I used to build an installation almost identical to mine, the same devices (router switch), almost identical CAPsMan configuration and version 6.48.1 and the problem does not occur at all. 33 days uptime and nothing has disappeared from the bridge. It is strange and interesting.
