Possible causes:

1. poor network connection between capsman and caps
2. high cpu load of capsman

If APs placed in same location, use load balancing group for distribution between balanced APs.

Does not matter, as it is called, (management or control)) the result is the same: cAPs select different channels to avoid neighbours interference.

2011 can serve up to 70 cAPs in capsman forwarding mode and up to 400 clients at same time, with 40-60% CPU load. Download speed make no matter, if capsman is a simple bridge (routing, queuing, firewalling ets served by another hardware), overwise you should keep in mind this workload.

Create hotspot on CAPsMAN, with capsman bridge as hotspot interface, will be fine )

Yes, CAPsMAN use DFS channel management by default. Some configuration tweaks for channel selection restriction may needed.

Seems CAPsMAN is single-threaded app, so degrade while connect-disconnect rate utilize single core performance. Here is two ways to avoid CAP disconnects: 1. Rise up single core frequency: use 3011 or CHR as single controller. 2. Set up multiple controllers: 2011 can serve 300-400 concurrent connect...

Has anyone used the CCR for hotspot deployments? I'm curious about max user counts and CPU load. Have test hotspot installation with 1036 as hotspot and 27 AP (SxT-G2HnD). Work fine, 70-150 online users (more than 4k in radius database) with simple queues. CPU usage less than 1-2%, peak traffic 30M...

I often give installers the device with a wire inserted in such a way.
Аdapter can not be lost, but the installer knows how to arrange the wire, so you laughter are not very relevant.

Any radius server + any billing system + some custom scripts)))

Simpliest way is to use Mikrotik UserManager

Hello all,
CCR 1036 RoS 6.3,
/ip hotspot cookie list contain only 4096 entries, this is bug or feature?

Same problem found on mipsbe v5.24.

I can't fetch data from mt box by ssh and cron because no reply from ssh (((

ssh -v look like.

Hello)

What about getting snmp counters of CRC/alinement errors on eth ports?
This information needed for automation of problem monitoring and preventive maintenance.

Normis, seems bottle neck is not queue but IRQ balancing|sharing for core ethernet. While reading forum, i often look complains about poor performance not only in queues and pptp tunnels, but also in general applications such VLANs. Also, according profiler, seems queue process utilize much less CPU...

Normis, same situation, as with OlegM. I tried to use ccr1036 as pptp NAS with queues and nat. Seems, when traffic come to around 100Mbps, some cores bring 100% use by IRQs, that may lead to lags and traffic forwarding stop. Else processes in profile look like low CPU consumption. [pater@ccrvpn] > t...

Hi guys. I was faced with incorrect work with RSTP in bridge, which includes vlan interfaces. In this case the vlan interface is the uplink, RSTP packets from the upstream switch come through in a tagged form, but the response mikrotik sends through this interface in untagged form, which leads to th...

Hello. In 6rc10 is possible to add more than one Address List in ppp profile, but when i send same parameter from radius, string addresslist1,addresslist2 received as single word, that lead to adding one new address list with name "addresslist1,addresslist2" There is debug log: 09:02:58 ra...

As decribed in wiki use-service-tag=yes setting used in qinq trunks, for regular tagged wlans should be set to NO.

use-service-tag=no