I own one cAP AC (RBcAPGi-5acD2nD r2) and one cAP Lite (cAP L-2nD). They were working as independent APs until now (different SSIDs among them) with the well-known Mikrotik wireless coverage and performance issues. They are both running RouterOS 6.47.1.
As I live in a two storeys house, I decided to use CAPSMAN to use them together and roam between them as I am moving around the house. I am well familiar with CAPSMAN due to numerous installations I have made, some with over 20 APs, either in local forwarding or CAPSMAN mode.
As soon as I enabled CAPSMAN (local forwarding) in both APs, I immediately noticed the following:
The ARM-based AP ( ) which was running CAPSMAN manager, showed excessive memory leakage. It started with 34 MB free/128 MB total and within 10-15 minutes the available memory dropped below 10 MB and the AP restarted itself with an “out of memory” error. Mind that all unnecessary packages, even IPv6, have been disabled. On the other hand, the MIPS-based CAP lite showed a consistent behavior in terms of system resources.
Even CAPSMAN was sending exactly the same settings as the stand-alone operation, all non-mobile phones or laptops could not connect to the CAPSMAN-controlled (in local forwarding mode) interfaces! The mobile phones (Samsung S6, S7, S9) and the laptops (HP/Dell) connected seamlessly!
Regarding (2), as soon as I disabled CAPSMAN and returned to the stand-alone AP configuration, all devices that were unable to connect before automatically connected! As a second test, I swapped SSIDs between the two APs… even in this case, all devices automatically connected to the “new” AP. I do not know what kind of manipulation CAPSMAN does when the APs are bonded. I suspect it has to do with the wireless bit, as the devices that are unable to connect do not ever appear even for a second within CAPSMAN registration tables, as if they cannot connect to the CAPSMAN-controlled SSIDs themselves and has nothing to do with upper-layer operation (DHCP addresses etc.). In addition, this behavior appeared either on dynamically-assigned IP devices or on statically-assigned ones.
Has anyone experienced either of the two above, the dramatic memory leakage/consumption and devices that are unable to connect while under CAPSMAN but connect fine on stand-alone operation?
with the well-known Mikrotik wireless coverage and performance issues
Nonsense…it’s more difficult to configure MikroTik Wifi (and router as well), I don’t have any issues.
The wireless network is dependant on CAPsMAN running. Because of the out of memory exception it might have caused these problems. Though being a device, the cAP ac is not a router (and also not suitable for running CAPsMAN).
Still a bit puzzled why you changed to CAPsMAN. Because (to me) it would make more sense to focus on configuration to have a decent Wifi network:
set SSID identical on all devices and all bands
set security identical (WPA2/AES only, and same key)
set transmission power as low as possible (to have a decent coverage)
I mentioned that I have performed many Miktotik installations, most of them with 2-figure APs. This does not mean however that the wireless part of these products perform not as advertised. I do not want to get into details, there are many posts within this forum and elsewhere depicting the less-than-average performance most times, not as a result of poor configuration. I have noticed it myself, comparing the same settings between Mikrotik and another (not mentioned on purpose) manufacturer. Of course, one must take into account the pricing factor as well, as no-one should demand a 70-Euro device to perform equally to a 300-Euro one. Honestly, I do not want to go into that and the purpose of this post is not to start a debate on whether these products are good or bad, but to try and figure out technically why this behavior is experienced.
Now, after these clarifications, on the technical aspect:
Agreed. I enabled CAPSMAN on cAP to verify the proof-of-concept, before investing to another wired router unit to permanently run CAPSMAN. This, however, does not mean that the unit should run out of memory as fast and at this rate. Nevertheless, with these aside, I describe a case where the same devices cannot even connect to the CAPSMAN SSIDs but they do connect if CAPSMAN is stopped (well before the out-of-memory incident start to occur). And this happens to printers and other IoT devices and not on mobile phones and/or laptops that do connect seamlessly to either configurations.
There are no overlapping channels among the APs and the CAPSMAN configuration is exactly the same as the stand-alone one. The CAPSMAN solution was attempted because there is a plan to add 4-5 more APs around the property for IoT and other coverage issues and projects. Hence, client roaming between APs is a must.
If I understand from the above, you suggest a "network" of 5-6 APs, which have their transmission power turned down (thus diminishing the coverage), each one broadcasting the same SSIDs with the same security settings (among the APs) and on non-overlapping channels, with no CAPSMAN or other interoperation among the APs, correct?